a bayesian network application for the ...a bayesian network application for the prediction of human...

NATIONAL TECHNICAL UNIVERSITY OF ATHENS

SCHOOL OF NAVAL ARCHITECTURE AND MARINE ENGINEERING

LABORATORY FOR MARITIME TRANSPORT

A BAYESIAN NETWORK APPLICATION

FOR THE PREDICTION OF HUMAN

FATIGUE IN THE MARINE INDUSTRY

AUTHOR: NIKOLAOS A. VAGIAS

THESIS SUPERVISOR: NIKOLAOS P. VENTIKOS

OCTOBER 2010

A BAYESIAN NETWORK APPLICATION FOR THE PREDICTION OF HUMAN

FATIGUE IN THE MARINE INDUSTRY 2

TABLE OF CONTENTS

LIST OF FIGURES ............................................................................................................... 6

LIST OF TABLES ................................................................................................................ 9

AKNOWLEDGEMENTS ................................................................................................... 10

ΠΔΡΙΛΗΨΗ ........................................................................................................................ 11

SYNOPSIS .......................................................................................................................... 14

1 THESIS PURPOSE .............................................................................................................. 18

1.1 BACKGROUND ........................................................................................................... 18

1.2 GENERAL GOAL ........................................................................................................ 18

2 HUMAN FACTORS AND RELIABILITY......................................................................... 21

2.1 HUMAN FACTOR AND HUMAN ERROR ............................................................... 21

2.2 THE EMERGENCE OF HFE ....................................................................................... 24

2.3 HUMAN RELIABILITY ANALYSIS AND HUMAN ERROR PROBABILITY ...... 26

2.4 HUMAN FACTOR IN MARITIME TRANSPORTS .................................................. 27

2.5 FATIGUE AS AN IMPORTANT HUMAN FACTOR ................................................ 30

3. FATIGUE ............................................................................................................................ 33

3.1. WHAT IS FATIGUE IN TERMS OF PHYSIOLOGY ............................................... 33

3.1.1 PHYSICAL FATIGUE .......................................................................................... 35

3.1.2 MENTAL FATIGUE ............................................................................................. 35

3.1.3 CENTRAL FATIGUE ............................................................................................ 36

3.1.4. PERIPHERAL FATIGUE ..................................................................................... 36

3.2. FATIGUE IN MEDICAL TERMS .............................................................................. 37

3.3 FATIGUE RISK FACTORS ......................................................................................... 38

3.3.1 SLEEP & SLEEP QUALITY ................................................................................. 39

3.3.2 CIRCADIAN RHYTHMS ..................................................................................... 47

3.3.3 SECONDARY RISK FACTORS ........................................................................... 50

3.4. EFFECTS OF FATIGUE ............................................................................................. 54

3.4.1. WORK PERFORMANCE .................................................................................... 54

3.4.2 MICRO SLEEPS .................................................................................................... 55

3.4.3 HEALTH EFFECTS .............................................................................................. 56

3.5 FATIGUE-RELATED ACCIDENTS ........................................................................... 57

3.5.1 COSTS ASSOCIATED WITH FATIGUE ............................................................ 58

3.6 MEASURING FATIGUE ............................................................................................. 59

3.6.1 PHYSIOLOGICAL MEASURES .......................................................................... 60



3.6.2 BEHAVIORAL MEASURES ................................................................................ 61

3.6.3 SELF REPORT MEASURES ................................................................................ 62

3.6.4 PERFORMANCE MEASURES ............................................................................ 64

3.6.5 OTHER FATIGUE MEASURMENTS TOOLS .................................................... 65

3.7 FATIGUE COUNTERMEASURES ............................................................................. 67

4 FATIGUE IN TRANSPORTATION ................................................................................... 72

4.1 GENERAL OVERVIEW .............................................................................................. 72

4.1.1 FATIGUE IN AVIATION ..................................................................................... 73

4.1.2 RAILWAY OPERATIONS FATIGUE ................................................................. 74

4.1.3 FATIGUE IN DRIVING ........................................................................................ 75

4.2 LEGISLATION REGARDING FATIGUE IN TRANSPORT ..................................... 77

4.3. FATIGUE IN MARITIME TRANSPORT .................................................................. 79

4.3.1 MARITIME FATIGUE RISK FACTORS ............................................................. 80

4.3.2 THE IMPACT OF STRESS ON MARINER‘S FATIGUE ................................... 93

4.4 RESEARCH REGARDING FATIGUE IN MARITIME TRANSPORATION ........... 94

4.4.1 ITF SEAFARER FATIGUE: Wake up to the dangers (1997) ............................... 94

4.4.2 THE NEW ZEALAND MARITIME SAFETY ..................................................... 95

4.4.3 THE CARDIFF PROGRAMM .............................................................................. 95

4.4.4 MAIB BRIDGE WATCHKEEPING SAFETY STUDY (2004) ........................... 96

4.4.5 THE TNO REPORT (2005) ................................................................................... 96

4.4.6 NTUA‘S IDENTIFYING AND ASSESSING NON-TECHNICAL SKILLS ON

GREEK MARITIME OFFICERS (2010) ....................................................................... 97

4.4.7 CREW ENDURANCE MANAGEMENT SYSTEM ............................................ 98

4.4.8 FATIGUE AT SEA - A FIELD STUDY IN SWEDISH SHIPPING..................... 99

4.5 LEGISLATION ON FATIGUE IN THE MARITIME SECTOR ............................... 100

4.5.1 IMO & THE ISM CODE ..................................................................................... 100

4.5.2 IMO AND STCW................................................................................................. 101

4.5.3 ILO 180 ................................................................................................................ 102

5 INTRODUCTION TO BAYESIAN NETWORKS ........................................................... 104

5.1 AN INTRODUCTION TO UNCERTAINTY AND GRAPHICAL MODELS .......... 104

5.1.1 UNCERTAINTY .................................................................................................. 104

5.1.2 PROBABILISTIC GRAPHICAL MODELS ....................................................... 105

5.2 BAYESIAN NETWORKS AND BAYESIAN INFERENCE .................................... 105

5.2.1 WHAT IS A BAYES NET? ................................................................................. 105

5.2.2 CONDITIONAL PROBABILITIES AND INDEPENDENCE IN BAYESIAN

NETWORKS ................................................................................................................. 107



5.2.3 BAYESIAN INFERENCE ................................................................................... 111

5.3 PRESENTATION OF A BAYESIAN NETWORK ................................................... 112

5.3.1 TYPICAL PRESANTATION OF A PROBLEM OF UNCERTAINTY ............. 112

5.3.2 USE OF BAYESIAN NETWORKS IN A TYPICAL EXAMPLE ..................... 113

5.3.3 USING GENIE ..................................................................................................... 116

5.3.4 NOISY OR GATES ............................................................................................. 120

5.4 COMMON BAYESIAN NETWORK APPLICATIONS ........................................... 121

5.4.1 GENERAL ........................................................................................................... 121

5.4.2 MARITIME .......................................................................................................... 123

5.4.3 FATIGUE ............................................................................................................. 123

5.5 WHY BAYESIAN NETWORKS ARE SO HELPFUL TO THESIS THESIS.

ADVANTAGES DISADVANTAGES ............................................................................. 124

6 A BAYESIAN NETWORK FOR PREDICTING FATIGUE IN MARITIME TRANSPORT

............................................................................................................................................... 128

6.1 MODEL PRESENTATION ........................................................................................ 128

6.1.1 TARGET NODE .................................................................................................. 128

6.1.2 PARENT NODES ................................................................................................ 129

7 BAYESIAN NETWORK APPLICATION FOR FATIGUE IN MARITIME SCENARIOS

AND ACCIDENTS ............................................................................................................... 155

7.1 THE IMPACT OF DYNAMIC FACTORS ON FATIGUE PREVALENCE ............ 156

7.1.1 SUFFICIENT SLEEP AND SLEEP DEPRIVATION......................................... 156

7.1.2 WEATHER AND SEA STATE ........................................................................... 158

7.2 MARITIME ACCIDENTS ......................................................................................... 161

7.2.1 EXXON VALDEZ ............................................................................................... 161

7.2.2 MFV BETTY JAMES .......................................................................................... 166

7.2.3 ANTARI (GENERAL CARGO 2446 G.T.) ........................................................ 172

7.3 HYPOTHETICAL SCENARIO OF 30000 G.T. CONTAINERSHIP ........................ 176

7.3.1 PROPOSED EVIDENCE ..................................................................................... 178

7.3.2 CALCULATING THE PROBABILITY FOR THE HYPOTHETICAL

SCENARIO ................................................................................................................... 180

8 SENSITIVITY ANALYSIS ............................................................................................... 185

8.1 INTRODUCTION ....................................................................................................... 185

8.2. BAYESIAN NETWORK SENSITIVITY ANALYSIS ............................................. 185

8.2.1 PARAMETER DESICION .................................................................................. 185

8.2.2 SENSITIVITY ANALYSIS ................................................................................. 187

8.2.3 NAVAL ARCHITECT OPTIMAL AND WORST SCENARIO ........................ 192

9 CONCLUSIONS ................................................................................................................ 197



9.1 GENERAL LITERATURE ......................................................................................... 197

9.2 CONCLUSION FROM OUR BAYESIAN NETWORK APPLICATION ON

MARITIME SCENARIOS ................................................................................................ 200

9.3 SUGGESTIONS .......................................................................................................... 202

REFERENCES .................................................................................................................. 204



LIST OF FIGURES

Chapter 2

Figure 2. 1: Types of Human Error .............................................................................. 22

Figure 2. 2: Three steps to Human Error avoidance .................................................... 23

Figure 2. 3: The Swiss Cheese Model for Human Error.............................................. 24

Figure 2. 4: A common example of Applied Ergonomics ........................................... 25

Figure 2. 5: Human Factors Engineering Pyramid ...................................................... 26

Figure 2. 6: The disasters of Exxon Valdez and Herald of Sea Enterprise .................. 27

Figure 2. 7: Maritime Human Factor Statistics ............................................................ 28

Chapter 3

Figure 3. 1: The Chernobyl nuclear accident ............................................................... 35

Figure 3. 2: Central and Peripheral fatigue .................................................................. 37

Figure 3. 3: Muscle impairment and rejuvenation process .......................................... 38

Figure 3. 4: The effect of sleep deprivation on alertness,(Sirois, 1998.) ..................... 41

Figure 3. 5: Average hours of sleep needed per day-Age ............................................ 42

Figure 3. 6: Examples of hazardous and healthy sleeping environments .................. 44

Figure 3. 7: Body temperature during a regular day (Sirois, 1998) ............................. 45

Figure 3. 8: Sleep deprivation and alertness (Sirois, 1998) ......................................... 46

Figure 3. 9: Biological clock and Circadian Rhythms ................................................. 48

Figure 3. 10: Circadian Rhythms and Alertness .......................................................... 48

Figure 3. 11: Fatigue-fitness connection (Ruby & Gaskill, 2002) ............................... 51

Figure 3. 12: Driver experiencing a microsleep........................................................... 56

Figure 3. 13: Three Mile Island nuclear power plant................................................... 59

Figure 3. 14: The Multiple Sleep Latency Test ........................................................... 60

Figure 3. 15: Fatigue measurement tools. PERCLOS and EEG ................................ 61

Figure 3. 16: A common behavioral measure, the Actigraph ...................................... 62

Figure 3. 17: The Epworth Sleeping Scale for sleepiness measurement ..................... 63

Figure 3. 18: A typical sleeping diary .......................................................................... 64

Figure 3. 19: Fatigue Severity Scale, (FSS) ................................................................. 67

Figure 3. 20: The Fatigue Avoidance Scheduling Tool ............................................... 69

Figure 3. 21: Effectiveness, feasibility and duration of effect for common Fatigue

Countermeasures .......................................................................................................... 70

Chapter 4

Figure 4. 1: The air crash of American Airlines, MD-82 at Little Rock, AR .............. 74

Figure 4. 2: A driver about to experience a microsleep ............................................... 76

Figure 4. 3: The first, most significant and most recently revised regulations ............ 77

Figure 4. 4: Legislation regarding Fatigue in Transportation ..................................... 79

Figure 4. 5: Sleep difficulties at sea, ashore and at home. ........................................... 82



Figure 4. 6: Vigilance though improved ship's design ................................................. 85

Figure 4. 7: The lack of sunlight in a vessel's Engine Room ....................................... 86

Figure 4. 8: ABS lighting Guidelines .......................................................................... 87

Figure 4. 9: ABS Guidelines regarding Noise levels aboard ships .............................. 89

Figure 4. 10: Transmission of vibration throughout the human body ......................... 90

Figure 4. 11: Vibrations effects to human body and movement .................................. 92

Figure 4. 12: Factors that affect the Marine's endurance (CEM) ................................. 98

Figure 4. 13: Factors that affect Fatigue according to Crew Endurance Management 99

Figure 4. 14: The Karolinska Sleepiness Scale .......................................................... 100

Chapter 5

Figure 5. 1: The difference between non-acyclic and acyclic graphs ...................... 106

Figure 5. 2: A casual network and its Conditional Probabilities Table ..................... 107

Figure 5. 3: Serial Connection ................................................................................... 109

Figure 5. 4: Diverging Connection ............................................................................ 109

Figure 5. 5: Converging Connection .......................................................................... 109

Figure 5. 6: Graphical presentation of the 'wet grass' example ................................. 112

Figure 5. 7: Noisy OR Gates example ....................................................................... 120

Chapter 7

Figure 7. 1: Fatigue Acceptance Scale ....................................................................... 155

Figure 7. 2: Fatigue - Hours Awake ........................................................................... 157

Figure 7. 3: The Beaufort Scale ................................................................................. 159

Figure 7. 4: Fatigue - Weather ................................................................................... 160

Figure 7. 5: The Exxon Valdez oil spill ..................................................................... 161

Figure 7. 6: Exxon Valdez actual incident ................................................................. 164

Figure 7. 7: Exxon Valdez alternate scenario ............................................................ 165

Figure 7. 8: Exxon Valdez scenario comparison ....................................................... 166

Figure 7. 9: The grounding of Betty James ................................................................ 167

Figure 7. 10: Betty James actual incident .................................................................. 169

Figure 7. 11: Betty James alternate scenario ............................................................. 171

Figure 7. 12: Betty James scenario comparison......................................................... 171

Figure 7. 13: The Antari Grounding .......................................................................... 172

Figure 7. 14: Antari actual incident ........................................................................... 174

Figure 7. 15: Antari alternate scenario ....................................................................... 176

Figure 7. 16: Antari scenario comparison .................................................................. 176

Figure 7. 17: A unfortunate incident aboard a containership.................................... 177

Figure 7. 18: Containership accident friendly scenario ............................................. 181

Figure 7. 19: Containership optimal scenario ............................................................ 182

Figure 7. 20: Containership scenario comparison...................................................... 183



Chapter 8

Figure 8. 1: Noise sensitivity analysis ....................................................................... 188

Figure 8. 2: Light sensitivity analysis ........................................................................ 189

Figure 8. 3: Vessels Motions sensitivity analysis ...................................................... 190

Figure 8. 4: Vibrations sensitivity analysis ................................................................ 191

Figure 8. 5: Management Demands sensitivity analysis ............................................ 192

Figure 8. 6: Naval Architect optimal scenario ........................................................... 193

Figure 8. 7: Naval Architect worst scenario .............................................................. 194

Figure 8. 8: Naval Architect scenario comparison ..................................................... 195



LIST OF TABLES

Chapter 3

Table 3. 1: Factors that affect the overall sleep quality ............................................... 40

Table 3. 2: Sleep architecture stages ............................................................................ 42

Chapter 5

Table 5. 1: P (W/R) and P (H/R, S) ........................................................................... 113

Table 5. 2: Prior probability table for P (H,R,S) ........................................................ 114

Table 5. 3: P (H,R,S) with entries H= n eliminated ................................................... 114

Table 5. 4: Calculation of P*( H,R,S) = P(H,R,S/ H=y) ............................................ 114

Table 5. 5: Calculation of P*(W,R)= R(W/R)*P*(R) =P(W,R) P*(R)/ P(R)

Table 5. 6: P**(W,R) =P( W,R /W=y, H=y) ............................................................. 115

Table 5. 7: P**(R,S) =P( R,S /H=y,W=y) ................................................................. 115

Chapter 7

Table 7. 1: Fatigue - Hours Awake ............................................................................ 157

Table 7. 2 : Fatigue -Weather .................................................................................... 160

Table 7. 3: Exxon Valdez Actual ............................................................................... 163

Table 7. 4: Exxon Valdez Alternate scenario ............................................................ 165

Table 7. 5: Betty James Actual incident .................................................................... 169

Table 7. 6: Betty James Alternate scenario ................................................................ 170

Table 7. 7: Antari actual incident ............................................................................... 174

Table 7. 8: Antari alternate scenario .......................................................................... 175

Table 7. 9: Containership accident friendly scenario................................................. 180

Table 7. 10: Containership optimal scenario ............................................................. 182

Chapter 8

Table 8. 1: Noise sensitivity analysis ......................................................................... 187

Table 8. 2: Light sensitivity analysis ......................................................................... 188

Table 8. 3: Vessel Motions sensitivity analysis ......................................................... 189

Table 8. 4: Vibrations sensitivity analysis ................................................................. 190

Table 8. 5: Management Demands sensitivity analysis ............................................. 191

Table 8. 6: Naval Architect Optimal scenario ........................................................... 193

Table 8. 7: Naval Architect worst scenario ................................................................ 194



AKNOWLEDGEMENTS

At this point, I feel obliged to mention several people who contributed, directly or

indirectly, to the completion of this thesis as well as the fulfillment of my academic

obligations for the School of Naval Architecture and Marine Engineering of the

National Technical University of Athens. For their significant assistance, guidance

and general support, I would like to thank the following:

Assistant Professor Nikolaos P. Ventikos, who served as this thesis supervisor, for his

invaluable assistance, patience and for giving me the opportunity to expand my

knowledge in the scientific domain that I wish to work in the future.

The entire faculty of the School of NAME, for their support regarding both academic

and personal matters.

My childhood friends, fellow NTUA students and especially Demi Rigou and Nikos

Piperis for reading this thesis and providing helpful comments and suggestions.

Mr. George Lycos, for giving me important directions at an early stage of my

endeavor.

Various scholars, students and professionals around the globe who were kind enough

to answer my questions and provide their expertise, experience and insight.

Finally and above all, I would like to thank my family for supporting me throughout

my entire life and helping to accomplish my goals and dreams.

Vagias Nikos

October 2010



ΠΕΡΙΛΗΨΗ

Σα ηειεπηαία ρξόληα ε επηζηήκε ηνπ αλζξώπηλνπ παξάγνληα έρεη μεθηλήζεη λα θάλεη

ζηγά ζηγά ηελ είζνδν ηεο ζην ρώξν ηεο παγθόζκηαο λαπηηιίαο. Παξόιν πνπ από

άπνςε ηερλνγλσζίαο θαη έξεπλαο βξίζθεηαη αξθεηά πίζσ από άιιεο πξνεγκέλεο

βηνκεραλίεο, νη εθαξκνγέο ηνπ θάλνπλ κηθξά αιιά ζηαζεξά βήκαηα πξνο ηελ

εδξαίσζε ηνπο ζην ρώξν. Ήδε έξεπλεο θαη ζηαηηζηηθά δεδνκέλα δείρλνπλ πσο ην

80% ησλ λαπηηθώλ αηπρεκάησλ είλαη ζπλδεδεκέλα, είηε έκκεζα είηε άκεζα, κε ην

αλζξώπηλν ζθάικα.

Έλα από ηα ζεκαληηθόηεξα ζηνηρεηά κε ηα νπνία ζρεηίδεηαη ην αλζξώπηλν ιάζνο είλαη

αλακθηζβήηεηα ε αλζξώπηλε απόδνζε. Με ηνλ όξν απηό ελλννύκε ην εύξνο ησλ

δπλαηνηήησλ ελόο αηόκνπ, θαζώο θαη ησλ πεξηνξηζκώλ πνπ ζέηεη ε αλζξώπηλε

θπζηνινγία ζηα πιαίζηα ηεο δηεθπεξαίσζεο ελόο θαζήθνληνο. Δίλαη ζαθέο πσο ν

αλζξώπηλνο νξγαληζκόο είλαη ζρεδηαζκέλνο κε ηέηνην ηξόπν έηζη ώζηε λα ακύλεηαη

ζε πεξίπησζε πνπ ηα παξαπάλσ όξηα μεπεξληνύληαη.

Η θόπσζε απνηειεί ην κήλπκα ηνπ αλζξώπηλνπ ζώκαηνο όηαλ ηα όξηα απηά έρνπλ

μεπεξαζηεί, είηε ηείλνπλ λα μεπεξαζηνύλ. Απνηειεί κηα θαηάζηαζε θαηάπησζεο θαηά

ηελ νπνία πεξηνξίδεηαη κεηαμύ άιισλ, ε ηθαλόηεηα ηεο αληίιεςεο, ηεο πξνζνρήο, ηεο

ζσκαηηθήο θίλεζεο θαζώο θαη ηεο αληίδξαζεο ελόο αηόκνπ ζε θπζηθά εξεζίζκαηα. Η

αληίδξαζε ηνπ ζώκαηνο ζηελ θόπσζε είλαη κηα παξαηεηακέλε πεξίνδνο μεθνύξαζεο

ε νπνία είλαη απαξαίηεηε ώζηε ε ιεηηνπξγηθόηεηα ηνπ νξγαληζκνύ λα επαλέιζεη ζε

θπζηνινγηθά επίπεδα θαη νπνηαδήπνηε δξαζηεξηόηεηα λα είλαη εθηθηή.

Δίλαη ζαθέο ινηπόλ πσο ε θόπσζε απνηειεί αλαπόζπαζην θνκκάηη ηεο αλζξώπηλεο

απόδνζεο θαη θαη‘ επέθηαζε ηεο επηζηήκεο πνπ κειεηά ν αλζξώπηλνο παξάγνληαο.

Μεξηθέο από ηηο κεγαιύηεξεο θαηαζηξνθέο ζηελ ηζηνξία ηεο αλζξσπόηεηαο ήηαλ

άκεζα ε έκκεζα ζπλδεδεκέλεο κε ην απνηέιεζκα κηαο ζεηξάο αηπρώλ ζπκβάλησλ πνπ

μεθίλεζαλ από έλα αλζξώπηλν ζθάικα. ε πνιιέο από απηέο ηηο πεξηπηώζεηο δε, ηα

αηπρήκαηα πξνέθπςαλ από ηηο πξάμεηο θαη ηηο επηινγέο ελόο αηόκνπ πνπ δξνύζε ππό

ηελ επήξεηα ηεο θόπσζεο. Ο θόζκνο αθόκα ζπκάηαη ηνλ ππξεληθό όιεζξν ηνπ

Σζεξλνκπίι, ηελ ηεξαζηία νηθνινγηθή θαηαζηξνθή πνπ πξνθάιεζε ε πξνζάξαμε ηνπ

Exxon Valdez θαζώο θαη ηα ακέηξεηα αεξνπνξηθά δπζηπρήκαηα πνπ θόζηηζαλ ηε

δσή ζε εθαηνληάδεο αλζξώπνπο.

ην πιαίζην απηήο ηεο δηπισκαηηθήο εξγαζίαο ζα εμεηάζνπκε ην θαηλόκελν ηεο

θόπσζεο ζηα λαπηηθά πιεξώκαηα, θαζώο θαη ηελ επίδξαζε απηήο ζε αηπρήκαηα

πινίσλ. ηε ζπλερεία, ζα παξνπζηάζνπκε έλα πηζαλνινγηθό κνληέιν βαζηζκέλν ζηε

ζεσξία πηζαλνηήησλ ηνπ Bayes, θαηαζθεπαζκέλν ππό ηε κνξθή ελόο δηθηύνπ

πίζηεσο ην νπνίν εθηίκα ηελ πηζαλόηεηα ηεο θόπσζεο ελόο λαπηηθνύ αλάινγα κε ην

εξγαζηαθό πεξηβάιινλ ηνπ, ηε θύζε ηνπ θαζήθνληνο ηνπ, ηε θπζηθή ηνπ θαηάζηαζε

θαη θπξίσο ηελ πνηόηεηα ηεο μεθνύξαζεο ηνπ νξγαληζκνύ ηνπ (θπξίσο πνηόηεηα

ύπλνπ) πάλσ ζην πινίν.



Σα δίθηπα πίζηεσο ή Bayesian Networks παξέρνπλ έλα ζύγρξνλν εξγαιείν γηα ηελ

εθηίκεζε πηζαλνηήησλ πνιύπινθσλ ζπζηεκάησλ ζε πεξηβάιινλ αβεβαηόηεηαο.

Απνηεινύλ έλα ζπλδπαζκό ηεο θιαζζηθήο ζεσξίαο πηζαλνηήησλ ηνπ Bayes θαη ηεο

ζεσξίαο γξαθώλ, αληιώληαο πιενλεθηήκαηα θαη δπλαηόηεηεο θαη από ηνπο δπν

ρώξνπο. ηηο κέξεο καο παξά πνιινί επηζηεκνληθνί θιάδνη έρνπλ επσθειεζεί από ηηο

εθαξκνγέο ηνπο κε θπξηόηεξα παξαδείγκαηα απηώλ ηελ ηαηξηθή θαη ηελ πιεξνθνξηθή.

Έλα απιό αληηπξνζσπεπηηθό παξάδεηγκα ησλ εθαξκνγώλ ηνπο είλαη ε εθηίκεζε ηεο

πηζαλόηεηαο ηεο εκθάληζεο κηαο λόζνπ δεδνκέλνπ ησλ πηζαλνηήησλ κηαο ζεηξάο

ζπκπησκάησλ ηεο.

ην 1ν θεθάιαην απηήο ηεο εξγαζίαο παξαζέηνπκε ηνπο ινγνύο πνπ καο νδήγεζαλ

ζην λα αζρνιεζνύκε κε ην ζπγθεθξηκέλν αληηθείκελν θαζώο θαη ην γεληθόηεξν ζθνπό

απηήο. Έπεηηα ζην 2ν θεθάιαην θάλνπκε κηα ζύληνκε αλαδξνκή ζηελ ηζηνξία ηνπ

αλζξώπηλνπ παξάγνληα θαζώο θαη ηεο εθαξκνγήο ηνπ ζην ρώξν ησλ κεηαθνξώλ

κέζσ ηνπ Human Factors Engineering. Αξγόηεξα, γίλεηαη αλαθνξά ζηελ αλάιπζε

αλζξώπηλεο αμηνπηζηίαο (Human Reliability Analysis) πνπ απνηειεί ην βαζηθόηεξν

εξγαιείν ηνπ αλζξώπηλνπ παξάγνληα γηα ηελ εθηίκεζε ησλ ζθαικάησλ. Σν θεθάιαην

νινθιεξώλεηαη κε κηα ζύληνκε αλαζθόπεζε ηεο εξγνλνκίαο ζην ρώξν ησλ

ζαιαζζίσλ κεηαθνξώλ ηνλίδνληαο ηε ζεκαληηθόηεηα ηνπ θαηλνκέλνπ ηεο θόπσζεο

ησλ λαπηηθώλ πιεξσκάησλ.

Σν 3ν θεθάιαην απνηειεί κηα εθηεηακέλε βηβιηνγξαθηθή επηζθόπεζε πάλσ ζην

γεληθόηεξν πιαίζην ηνπ θαηλνκέλνπ ηεο θόπσζεο σο θνκκάηη ηεο επξύηεξεο

αλζξώπηλεο θπζηνινγίαο. Παξάγνληεο όπσο ν ύπλνο, νη θηξθάδηνη ξπζκνί θαη ην ζηξεο

πνπ νδεγνύλ ζηελ θόπσζε αλαιύνληαη ζε βάζνο ελώ ηειηθά γίλεηαη αλαθνξά ζηηο

ζπλέπεηεο ηεο, ζε κέηξα αληηκεηώπηζεο ηεο θαζώο θαη ζε επηζηεκνληθνύο ηξόπνπο

κέηξεζεο θαη πνζνηηθνπνίεζεο ηεο.

ηε ζπλερεία, ην 4ν θεθάιαην αλαθέξεηαη απνθιεηζηηθά ζηε παξνπζία θαηλνκέλσλ

θόπσζεο ζην ρώξν ησλ κεηαθνξώλ. Αθνύ θάλνπκε κηα ζύληνκε αλαθνξά ηνπ

αληηθείκελνπ ζε άιιεο βηνκεραλίεο, όπσο νη αεξνκεηαθνξέο θαη νη νδηθέο

ζπγθνηλσλίεο, παξαζέηνπκε ηνπ παξάγνληεο πνπ πξνθαινύλ θόπσζε ζηα λαπηηθά

πιεξώκαηα ζπκθώλα κε ηνλ IMO. Ιδηαίηεξε έκθαζε δίλεηαη ζε εθείλνπο ηνπο

παξάγνληεο κέζσ ησλ νπνίσλ ν Ναππεγόο Μεραλνιόγνο Μεραληθόο κπνξεί λα

δεκηνπξγήζεη έλα πγηέο θαη εξγνλνκηθό πεξηβάιινλ, απαιιαγκέλν από ηηο βιαβεξέο

επηπηώζεηο ηεο θνύξαζεο ησλ πιεξσκάησλ. Σέινο, γίλεηαη κηα ζύληνκε

βηβιηνγξαθηθή επηζθόπεζε ζε αληίζηνηρεο έξεπλεο πνπ έρνπλ γίλεη ζην παξειζόλ ζην

ρώξν ηεο λαπηηιίαο θαζώο θαη ζηνπο ππάξρνληεο ζεζκνζεηεκέλνπο θαλνληζκνύο θαη

λνκνζεζίεο.

Σν 5ν θεθάιαην απνηειεί κηα κίλη-εηζαγσγή ζηελ αβεβαηόηεηα, ηα πηζαλνινγηθά

γξαθηθά κνληέια θαη ηηο βαζηθέο αξρέο ηεο ζεσξίαο πηζαλνηήησλ. Γίλνληαη

αληηπξνζσπεπηηθά παξαδείγκαηα, κέζσ ησλ νπνίσλ γίλεηαη επεμήγεζε ηνπ νξηζκνύ

ελόο δηθηύνπ πίζηεσο, ελώ κε ηε ρξήζε ελόο κίλη-tutorial πξνζπαζνύκε κε ζπλνπηηθό

ηξόπν λα δείμνπκε ηηο βαζηθόηεξεο ιεηηνπξγίεο θαη δπλαηόηεηεο ηνπ ινγηζκηθνύ πνπ



ρξεζηκνπνηήζεθε. ηε ζπλέρεηα παξαζέηνληαο ηα θπξηόηεξα πιενλεθηήκαηα ησλ

δηθηύσλ πίζηεσο, εμεγνύκε γηα πνηνπο ιόγνπο θαηαιήμακε ζηελ επηινγή ηνπο γηα ηελ

εξγαζία καο. Σειηθά, ζην θεθάιαην πνπ αθνινπζεί παξνπζηάδνπκε αλαιπηηθά ηε

δνκή ηνπ κνληέινπ καο, παξαζέηνληαο ηνπο θόκβνπο πνπ ην απνηεινύλ, ηηο ζρέζεηο

κεηαμύ απηώλ (ππό ηε κνξθή πηλάθσλ εμαξηεκέλσλ πηζαλνηήησλ) θαζώο θαη ηηο

αξρηθέο θαηαλνκέο ησλ πηζαλνηήησλ ησλ θόκβσλ-γνλέσλ.

ην 7ν θεθάιαην, ην νπνίν απνηειεί ην βαζηθό κέξνο ηεο εξγαζίαο καο, γίλεηαη

εθαξκνγή ηνπ πηζαλνινγηθνύ δηθηύνπ καο ζε πξαγκαηηθά λαπηηθά αηπρήκαηα, αιιά

θαη ζε ξεαιηζηηθά ππνζεηηθά ζελάξηα. Η αλαλέσζε ηνπ δηθηύνπ γηα ηνλ ππνινγηζκό

ηεο ηειηθήο πηζαλόηεηαο ηεο θόπσζεο γίλεηαη κέζσ ηεο εηζαγσγήο απνδείμεσλ. Οη

απνδείμεηο απηέο απνηεινύλ πξντόλ είηε επηζεσξήζεσλ λαπηηθώλ αηπρεκάησλ ππό ηε

κνξθή γξαπηώλ αλαθνξώλ, είηε απνηειέζκαηα εθηηκήζεσλ πνπ πξνήξζαλ από

πξαγκαηηθέο καξηπξίεο λαπηηθώλ αιιά θαη ηελ εθηεηακέλε έξεπλά καο πάλσ ζην

ζέκα. Από ηελ παξαπάλσ δηαδηθαζία πξνέθπςαλ αληηπξνζσπεπηηθά δηαγξάκκαηα θαη

πηλάθεο, ηα νπνία κε ηε ζεηξά ηνπο βνήζεζαλ ζηελ θαιύηεξε θαηαλόεζε ηνπ

εμεηαδόκελνπ ζέκαηνο.

ην 8ν θεθάιαην πξαγκαηνπνηήζακε κηα αλάιπζε επαηζζεζίαο (Sensitivity Analysis)

κε ζθνπό λα ειέγμνπκε ηελ γεληθόηεξε αμηνπηζηία θαη ζπκπεξηθνξά ηνπ κνληέινπ

καο θαζώο θαη ην θαηά πόζν ηα απνηειέζκαηα αληαλαθινύλ ηελ πξαγκαηηθόηεηα.

Έπεηηα, ζειήζακε λα δώζνπκε κηα ξεαιηζηηθή εθηίκεζε ηνπ πνζνζηνύ πνπ κπνξεί λα

ζπλεηζθέξεη ν Ναππεγόο Μεραλνιόγνο Μεραληθόο ζηελ βειηίσζε ησλ ζπλζεθώλ

δηαβίσζεο πάλσ ζην πινίν θαη θαη‘ επέθηαζε ζηελ αληηκεηώπηζε θαηλνκέλσλ

θόπσζεο ζηα λαπηηθά πιεξώκαηα.

Κιείλνληαο, ε εξγαζία καο παξνπζηάδεη ηα θπξηόηεξα επξήκαηα πνπ πξνέθπςαλ ηόζν

από ηελ βηβιηνγξαθία όζν θαη από ηελ εθαξκνγή ηνπ κνληέινπ. Έπεηηα, παξαζέηνπκε

ηα ζεκαληηθόηεξα ζπκπεξάζκαηα καο πάλσ ζην ζέκα ζε αληηζηνηρία κε πξνηάζεηο

πνπ πηζηεύνπκε όηη ζα νδεγήζνπλ ζηελ απνηειεζκαηηθόηεξε θαηαπνιέκεζε

θαηλνκέλσλ θόπσζεο ζηηο ζαιάζζηεο κεηαθνξέο.



SYNOPSIS

In the past few years, human factors science has begun to make steps towards its

entrenchment in the global maritime domain. Even if the shipping world is at an early

stage in terms of technical know-how and research, compared to other advanced

industries, human factors applications are in the process of their consolidation in the

field. Both researches and statistical data show that 80% of total maritime accidents

are connected, indirectly or directly, to the human fault.

One of the fundamental elements related to the human error is undeniably an

individual‘s performance. By using this term we define the breadth of possibilities of

an individual as well as the limitations that places human physiology in terms of

performing of duty. It is explicit to indicate that the human body is structured in such

way that defense mechanisms are activated when the preceding limits are exceeded.

Fatigue constitutes the signal of the human body that serves as an alert when the

bodily limitations are surpassed or tend to be exceeded. It is describe by medicine, as

a condition of collapse during which awareness, attention, bodily movement and

reaction time to stimuli are severely diminished. The organism‘s answer to fatigue is

the need for an extended period of rest which is essential for the human body in order

for any ordinary activity to be feasible in the future.

It is explicit therefore that fatigue constitutes an integral factor of human performance

thus being an extension of science that studies human factors. Some of the notable

dissasters in the history of humanity were directly or indirectly connected with the

sequence of mishaps that began from a human fault. In many of these cases, the

accidents resulted from the actions and the choices of a certain individual that acted

under the influence of fatigue. The world still remembers the nuclear havoc of

Chernobyl, the enormous ecological destruction that caused the grounding of Exxon

Valdez as well as the innumerable aviation accidents that cost the life of hundreds of

people.

In this thesis, we examine the phenomenon of fatigue amongst shipping crews as well

as its effect on maritime accidents. Later, we will present a probabilistic model based

on Bayesian probability theory, constructed under the form of a belief network which

assesses the mariner‘s fatigue, given information depending his work environment,

the nature of his duty, his physical condition and primarily the rest quality (mainly

sleep quality) that a seafarer has the chance to amass onboard.

Belief networks or Bayesian Networks, as they are commonly known, provide a state

of the art tool for the probabilities inference calculus of complex and sophisticated

systems, in an uncertainty environment. They constitute a marriage of traditional

Bayesian theory and Graph theory, drawing advantages and capabilities from both

fields. Nowadays many scientific domains have profited from their applications, with

medicine and the computer science being the most typical examples. A simple



representative example of their applications is the probability inference for the

prevalence of a disease, given the probabilities for a plethora of its symptoms.

In the 1st chapter of this study we indicate the reasons that led us to research this

particular subject, as well as the general purposes of this thesis. Then in the 2nd

chapter we make short reference in the history of human factors as well as its

incorporation in the transportation field via the implementation of Human Factors

Engineering. Later, we discuss the principles of Human Reliability Analysis (HRA)

which constitutes the fundamental tool that Human Factor science uses for the

prediction of the human error. Finally, we make a brief introduction to ergonomic

principles in maritime transport while we highlight the significance of fatigue

amongst seafarers.

The 3rd chapter is an extensive bibliographic review on the prevalence of fatigue as

part of general human physiology. Risk factors such as sleep, circadian rhythms and

stress, all of which lead to lassitude, are analyzed in-depth while references are made

regarding fatigue consequences, countermeasures as well as scientific tools for fatigue

measurement and quantification.

Furthermore, the 4th chapter is dedicated exclusively to fatigue in the field of global

transports. After a short review of fatigue prevalence in industries such as air

transportations and railways, we indicate the factors that cause fatigue in the maritime

domain in accordance to the guidelines provided by IMO. Particular emphasis is

given in those factors through which the Naval Architect and Marine Engineer can

create a healthy and ergonomic environment, exempted from the damaging

repercussions of tiredness amongst crews. Finally, we provide a brief review of

related researches and projects as well as of existing enacted regulations and

legislation.

The 5th chapter constitutes a mini-introduction to uncertainty, probabilistic graphic

models and the basic principles of probabilities theory. Though the use of

representative examples, the definition of Bayesian Networks is explained, while a

mini-tutorial is provided in order to demonstrate the basic functions and overall

capabilities of the GeNie (software used). Furthermore, the primary advantages of

Bayesian networks are listed in order to highlight the reasons that led us in their

choosing as the instrument of our research. Finally, in the next chapter we present the

structure of our model in depth, introducing the participating nodes, the relations

between them (in the form of conditional probabilities tables) as well as the initial

prior probabilities distributions of the parent-nodes.

In the 7th chapter, which constitutes the main part of our thesis, we apply our network

to real maritime accidents as well as in realistic hypothetical scenarios. Each time, we

import evidence to our network and through BN updating the posterior fatigue

probability is calculated. This evidence was product either of inspections of actual

accidents in the form of official reports, or results of estimations from seamen‘s‘



testimonies and our own extensive research on the subject. The preceding process

provided several representative figures and tables which in turn helped us in to

achieve better comprehension of fatigue in the shipping word.

In the 8th chapter we performed a Sensitivity Analysis in order to test the overall

reliability and behavior of our model as well as its reflection of the ‗real world‘. Then,

we try to give a realistic estimate of percentage of the Naval Architect‘s possible

contribution in the improvement of the living conditions aboard ships; a contribution

that can help can mitigate fatigue phenomena amongst seafarers.

Finally, we present our major findings that resulted both from our extensive

bibliographic review and the application of our model. In addition, we summarise our

conclusions on the subject in equivalence with proposals that we believe will lead into

more effective countermeasures and thorough knowledge about fatigue at sea.



CHAPTER 1 THESIS PURPOSE



1 THESIS PURPOSE

1.1 BACKGROUND

In the last 30 years, the maritime industry has suffered many losses, both in terms of

human lives and ecological disasters, caused by the hazardous effects of the

prevalence of fatigue aboard ships. Even if the principles of Human Factors are

becoming more and more incorporated into the shipping world, the occurrence of

fatigue amongst shipping crews is rather underrated and often neglected. Furthermore,

the contemporary tendency to reduce operating costs has resulted in lower manning

levels, exhaustive workloads and badly planned shift rotation. In addition, regular port

calls, adverse weather conditions and intensive paperwork lead mariners in adjusting

to a rather hazardous lifestyle which makes them vulnerable to the obnoxious

consequences of fatigue.

The consequences of fatigue amongst mariners lead in dangerous outcomes both in

terms of the vessel‘s overall safety but the individual‘s well being as well. Common

examples of fatigue include injuries that were the result of lack of alertness, or even

major ecological catastrophes, such as the grounding of the Exxon Valdez, that was

the outcome of the actions of a sleep deprived watchkeeper. Additionally, taking into

consideration the number of people aboard a regular passenger vessel, a possible

fatigue-related incident could result in the loss of thousands of human lives.

While searching existing literature, interesting but rather unsettling facts can be

found. Not only has there been relatively little research on the subject but the

mariner‘s knowledge regarding fatigue ranges from limited to none. It is clear that if

shipping crews were more aware of fatigue prevalence and its mechanisms, they

would have been more careful in their choices and actions aboard ships.

Consequently, fatigue perception and knowledge is an integral part in the effort of

mitigating fatigue related incidents in the maritime domain.

1.2 GENERAL GOAL

The general goal of this thesis can be summarized in the following paragraphs:

At first, we tried to provide a thorough literature framework regarding the subject of

fatigue and its overall nature. While circumventing on the physiological part of the

issue, we present the basic biological mechanisms of fatigue as well as the mental and

body-motor effects on a human organism. Furthermore, the most important factors

that lead to fatigue were highlighted and described in order to indicate their

significance on its prevalence. Through this procedure we acquired sufficient

knowledge which was later used in the structure of our probabilistic graphical model.



However, our main purpose in aggregating all this information is to grant the reader

the opportunity to develop fatigue perception and attain substantial knowledge about

the subject.

Later in this thesis, we propose the use of a probabilistic graphical model for the

prediction of fatigue, given evidence regarding the environment and the individual‘s

physiological background. The primary purpose of our model (and the thesis as well)

can be subcategorized in two main goals:

To provide the opportunity to predict the posterior fatigue probability in a real

maritime accident. Through this procedure we are able to identify the

importance of fatigue prevalence in an actual maritime mishap and the

significance of the contributing factors that led to it

To provide an efficient auditing tool that will help mariners and investigating

parties to evaluate fatigue in the future. Through proper use and further

improvement of our model surveyors can benefit from fatigue related findings,

such as extreme noise and vibrations levels that are rarely mentioned in an

actual accident report. The primary advantage of our model is that it provides

the opportunity to predict fatigue prevalence and its importance, given

information regarding workload, environment and ergonomic factors, prior to

the occurrence of the accident.



CHAPTER 2 HUMAN FACTORS AND

RELIABILITY



2 HUMAN FACTORS AND RELIABILITY

2.1 HUMAN FACTOR AND HUMAN ERROR

The concept of human error, whether intentional or unintentional, is defined as

(Lorenzo, 1990):

Any human action (or lack thereof) that exceeds or fails to achieve some limit of

acceptability, where limits of human performance are defined by the system.

Human error has been the main contributor in numerous modern disasters such as the

factory explosion in Bhopal, the railway accidents at Paddington and Southall, the

shipwrecks of Herald of Free Enterprise and Exxon Valdez, the nuclear catastrophes

in Chernobyl and in Three-Mile Island, the accident of the Challenger Shuttle and

many more. What these disasters had in common was that they were exposed at a

daily basis and finally affected by the impact of human factor/error. Many of these

incidents resulted in loss of life or multiple injuries but apart from that, the

environmental and monetary costs were even greater [108].

Managing and preventing human error has become a top priority for most of the

advanced industries such as aviation and nuclear. Initially, researchers focused on

the technical part of the problem, a common practice that has been dominant ever

since. However, since contemporary technology has provided extremely intelligent

and reliable systems, scientists also focused on the interplay between these systems

and human performance. Moreover, while investigating accidents that included

equipment and operator failure, researchers agree that there should be more available

knowledge (in terms of training) regarding safety culture.

The following are some of the many ways to categorize human error [I].

exogenous versus endogenous

situation assessment versus response planning and related distinctions in

o errors in problem detection

o errors in problem diagnosis (problem solving)

o errors in action planning and execution

By level of analysis; for example, perceptual versus cognitive versus

communicational versus organizational.

Figure 2.1 shows a logical sequence resulting in human error. It is obvious that human

error can occur in each one of these steps.

http://en.wikipedia.org/wiki/Problem_solving

http://en.wikipedia.org/wiki/Interpersonal_communication

http://en.wikipedia.org/wiki/Organizational_studies



Figure 2. 1: Types of Human Error

Recent statistical data prove that over 80% of the accidents that take place globally

are entirely or partly connected to human error or underperformance (HSE, 1999). For

example, a study has shown that the human error is the main contributor to 60% of

total maritime accidents, with the organizational and management omissions to

constitute 15%, while the rest 25% to be part of technical problems (Tangen, 1987).

In the search for mitigating the hazardous effects of human error, the field of Human

Factors came into prominence. According to Dr.Rothblum [80] the principle of

Human Factors is the instrument for understanding and managing human potential in

terms of capabilities and limitations. Human factors can prove helpful in applying its

knowledge to equipment, work environments, procedures, and policies in order to

enhance their interaction with human performance. By integrating applied ergonomics

in technology HF guarantees a type of informal „alliance‟ between individuals and

their overall working environment. The idea of adapting complex systems to human

performance has been widely described by scholars as the human-centered approach.

This practice can provide several beneficial elements to individuals such as increased

efficiency and effectiveness, reduced error probability, lower costs, decreased

personnel injuries and above all increased morale and well being [80].

Human factors science has provided applications for various domains such as

engineering, statistics, medicine, operations research and psychology. It is a term that

covers (Figure 2.2):

The science of understanding the properties of human capability (Human

Factors Science).

The application of this understanding to the design, development and

deployment of systems and services (Human Factors Engineering).

The art of ensuring successful application of Human Factors Engineering to a

program (Human Factors Integration). It can also be called ergonomics [I].

http://en.wikipedia.org/wiki/Engineering

http://en.wikipedia.org/wiki/Statistics

http://en.wikipedia.org/wiki/Operations_research

http://en.wikipedia.org/wiki/Anthropometry

http://en.wikipedia.org/wiki/Human_Factors_Integration



Figure 2. 2: Three steps to Human Error avoidance

Some more specializing approaches of Human Factors include:

Cognitive ergonomics

Usability

Human computer interaction

Human machine interaction

User experience engineering.

In the past 20 years Human Factors has been a very active research field. Specific

focus has been given in researching memory capacity. Another field that has also

benefited from the use of Human Factors practices is decision making in terms of

preventing human error probability. Furthermore, conversation analysis has used

several HF patterns in order to develop a productive human communications

framework.

In an attempt to formally classify various types of human error, Dr Scott Shappell and

Dr Doug Wiegmann of the University of Illinois, developed the Human Factors

Analysis and Classification System (HFACS). HFACS is a useful tool in identifying

the human causes of an accident and providing assistance in the investigation process,

target training and prevention efforts, (Kirwan and Ainsworth 1992). It is based in the

"Swiss Cheese" model of human error which looks at four levels of human failure,

including unsafe acts, preconditions for unsafe acts, unsafe supervision, and

organizational influences. It is a comprehensive human error framework that folded

http://en.wikipedia.org/wiki/Memory

http://en.wikipedia.org/wiki/Decision_making



Reason's ideas into the applied setting, defining 19 causal categories within four

levels of human failure [AA].

Figure 2.3 gives a visual representation of the Swiss Cheese Model.

Figure 2. 3: The Swiss Cheese Model for Human Error

2.2 THE EMERGENCE OF HFE

Human Factors science has been the spark for further and more multifaceted studies

regarding human potential and performance. In the wake of the 21st century, a new

engineering field was introduced in order to integrate Human Factors knowledge to

modern engineering.

“Human Factors Engineering (HFE) is the discipline of applying what is known

about human capabilities and limitations to the design of products, processes,

systems, and work environments. It can be applied to the design of all systems having

a human interface, including hardware and software. Its application to system design

improves ease of use, system performance and reliability, and user satisfaction, while

reducing operational errors, operator stress, training requirements, user fatigue, and

product liability. HFE is distinctive in being the only discipline that relates humans to

technology.” [I]



Human Factors engineering activities include:

1. Usability assurance

2. Determination of desired user profiles

3. Development of user documentation

4. Development of training programs [I].

An example of HFE integration is figure 2.4. Human Factors science and HFE

have pinpointed the parameters that affect directly a computer operator‘s

compliance to the system. In order for the operator to achieve maximum

performance, HFE has provided the optimum distances and angles ranges in a

casual workplace environment.

Figure 2. 4: A common example of Applied Ergonomics

Furthermore, Human Factors Engineering has provided guidelines in integrating its

principles in ship design as well. Some of the most common suggestions include [14]:

identifying the individuals part in a complex system

workload simulation and modeling

enhancement of man-machine interfaces in order to increase overall safety

advanced human centered ship design methods

techniques to enhance ship crew productivity thereby reducing manning

requirements.

In figure 2.5 HFE suggests, in the form of a pyramid, the most important steps for

successfully integrating its principles in modern industries. The bottom of the pyramid

points the fact that before any action is taken, the management has to realize that



following this procedure can be beneficial to the personnel but to the whole

company/industry as well.

Figure 2. 5: Human Factors Engineering Pyramid

2.3 HUMAN RELIABILITY ANALYSIS AND HUMAN ERROR

PROBABILITY

The most popular method for assessing human error probability in human/machine

environment is the Human Reliability Analysis (HRA), a state of the art tool based on

the principles of Probabilistic Risk Assessment (PRA). It can be best described as an

analysis of the potential failure arising from human performance which emphasizes

on system risk and reliability.

Human reliability analysis grew up in the 1960s with the intention of modeling the

likelihood and consequences of human error. Initially, humans were considered as

typical system component no different than equipment or procedures. The main part

of an HRA is the assessment of the operator‘s error probability. Over the years,

methods have been developed that recognize human potential to recover from a

failure, on the one hand, and the effects of stress and organizational culture on the

likelihood of possible errors, on the other. However, no method has yet been

developed that incorporates all our understanding of individual, team and

organizational behavior into overall assessments of system risk or reliability [2].

Some of the most popular HRA‘s include:



A Technique for Human Error Analysis ( ATHEANA)

Cognitive Reliability and Error Analysis Method (CREAM )

Human Cognitive Reliability Correlation (HCR )

Human Error Assessment and Reduction Technique (HEART)

Influence Diagrams Approach (IDA )

Success Likelihood Index Method (SLIM )

Technique for Human Error Rate Prediction (THERP)

2.4 HUMAN FACTOR IN MARITIME TRANSPORTS

In the past three decades, the maritime industry has suffered from incidents and

accidents that were associated directly or indirectly with the prevalence of human

error at sea. Maritime disasters such as the Exxon Valdez or the Herald of Sea

Enterprise led both the scientific world and various organizations (e.g. classification

societies) to conduct further investigation on the subject. While investigating the

reports of most of these incidents, mariners‘ fatigue, human error, and human

performance are recorded on a regular basis. International groups such as the

International Maritime Organization (IMO) have established new legislation and

proposed guidelines in order to address human error. Flag State safety, enforcement

agencies and international treaties have implemented new laws and proposed several

programs and strategies to reduce human error at sea. Moreover, high-risk ships can

now be identified while conducting port state controls practices at a regular basis (e.g.

mini bulkers).

Figure 2. 6: The disasters of Exxon Valdez and Herald of Sea Enterprise

Moreover, the whole maritime community, including IACS and national coast guards,

has tried to isolate specific risk factors that increase human error probability. This fact

has forced shipping companies to comply with existing regulations and embrace any

human error related guidelines [53].



Rothblum [80] states that the maritime system is a people system, and human errors

figure prominently in casualty situations. About 75-96% of marine casualties are

caused, at least in part, by some form of human error. Studies have shown that human

error contributes to:

• 84-88% of tanker accidents

• 79% of towing vessel groundings

• 89-96% of collisions

• 75% of allisions

• 75% of fires and explosions

Figure 2. 7: Maritime Human Factor Statistics [80]

Some typical examples of the measures taken to mitigate human error on board was

the establishment of the Integrated Bridge Systems (IBS), the improvement of the

vessel‘s habitability and several applications of applied ergonomic in the engine room

and the engine control room.

However, the most crucial measures to battle human error have been the introduction

of the International Safety Management code (ISM) and several amendments on the

Standards for Certification and Training of Watchkeepers (STCW) in the 2010 Manila

convention. The fact that over 80% of accidents at sea are directly or indirectly a

result of human error is primarily the reason for further research, investigation and

guidance.

Pomeroy and Sherwood Jones of Lloyd‘s Register of shipping [70] indicate that the

operational context is changing while the mariner population is changing in terms of

0

10

20

30

40

50

60

70

80

90

100

Human Factors Related Incidents

Tanker accidents

Groundings

Collisions

Allisions

Fires/Explosions



background, culture and skills. The need to reduce overall costs has forced shipping

companies to minimize crew complements. Moreover, seafarers have to adapt to high

tech equipment and complex systems, results of the contemporary technological

evolution and dominance of computer based machinery. While it is not obvious that

these independent developments are compatible to human capabilities, the shipping

industry remains in a transitional phase between a goal-setting approach and several

regulation based and prescriptive trends.

IMO, in order to formally address the human element in the maritime domain, has

established the IMO HUMAN ELEMENT GROUP. This group was formed in 1991 by

the fifty-ninth session of the Maritime Safety Committee (MSC) and the thirty-first

session of the Marine Environment Protection Committee (MEPC).

The fact that the working group was established jointly by both major IMO

committees, as listed above, is in recognition of the case that the human element is a

key factor in both safety and pollution prevention issues. The working group meets

concurrently with meetings of the MSC or MEPC, as necessary.

Nowadays the group has several separate parties that address each specialized task

through correspondence. Furthermore, the Human Element Working Group

collaborates directly with the group that deals with Formal Safety Assessment (FSA).

The FSA group aims to generate a new approach for developing international

regulations with emphasis on risk and hazard assessment. This approach also gives

full attention to the human element in ship operations.

The most important issues addressed by the Human Element Working Group include:

safety management, through the ISM Code;

human element principles and goals for the Organization;

human element analyzing process tool for addressing the human element in

the regulatory process;

the problems associated with fatigue;

a taxonomy of terms used in human element analysis; and

review of studies related to ship operations and management.

Consideration of the human element has become a standing subject within the

Organization. The Organization has adopted Human Element Vision, Principles and

Goals and has developed the Human Element Analyzing Process tool, which provided

a strategic plan and a structured systematic framework for addressing the human

element [K].

In her paper regarding maritime Human Factors, Dr. Rothblum [80] summarizes the

most common Human Factors Issues in the Marine Industry:

http://www.imo.org/Newsroom/index.asp?topic_id=110

http://www.imo.org/Newsroom/index.asp?topic_id=109

http://www.imo.org/Newsroom/contents.asp?doc_id=648&topic_id=227

http://www.imo.org/HumanElement/index.asp?topic_id=177

http://www.imo.org/HumanElement/index.asp?topic_id=177



Fatigue. 16% of the vessel casualties and 33% of the injuries are

fatigue related.

Inadequate Communications. 70% of major marine collisions and

allisions occurred while a State or federal pilot was directing one or

both vessels

Inadequate General Technical Knowledge. In one study, this problem

was responsible for 35% of casualties

Inadequate Knowledge of Own Ship Systems.The lack of ship-specific

knowledge was cited as a problem by 78% of the mariners surveyed.

Poor Design of Automation. Decisions Based on Inadequate

Information.

Faulty standards, policies, or practices

Poor maintenance

Hazardous natural environment.

2.5 FATIGUE AS AN IMPORTANT HUMAN FACTOR

In our search for significant human factors in the maritime domain we concluded that

one the most important ones is human physiology. It is rather logical that when a

piece of machinery is badly maintained the probability of the system‘s failure is

significantly high. The same rule occurs while investigating human performance.

Fatigue is the primary factor of human underperformance since it prevents the human

body and cognitive skills to perform at an optimum level. While it is generally

researched in the maritime industry, most studies have failed to indentify its

importance and long term effect that may come with it. The underreporting of fatigue

and the difficulties in measuring and analyzing it have made this task even harder.

According to Dr. Paula Sind-Pruiner of United States Coast Guard Research and

Development center [14] there are a number of factors that have been recognized

concerning human factors and fatigue in the maritime industry. Some of them are

listed below:

When stressed, fatigued, overworked, etc., trained methods usually fail.

Fatigue is alive and well in the maritime industry.



There are competitive economic pressures to maximize vessel utilization.

There are competitive economic pressures to minimize crew size.

Failure to recognize the dangerous performance effects of fatigue.

Lack of knowledge concerning fatigue, and

System lifecycle training costs often greatly exceed the incremental design

costs.[14]

Hethrighton et al. [31] state that contemporary research has shown that there are

potentially disastrous outcomes from fatigue in terms health and performance.

In the wake of the twenty first century, the seafarer‘s job has become more

demanding than ever due to a combination of minimal manning, sequences of rapid

turnarounds and short sea passages, adverse weather and traffic conditions, long

working hours with insufficient opportunities for recuperative rest [94].

Despite the introduction of work rest mandates by IMO (2001) and STCW (1995)

many seafarers have to work for more than 12 hours a day with a maximum of 5-6

hours break. In addition, in a report by the National Transportation Safety Board

(1999), seafarers were identified out of the occupational groups included to be second

in highest number of continuous working time, close behind rail workers.

Moreover, the seafarers themselves recognized that the impact of fatigue has been

significantly greater in the past 10 years than it used to be. In a certain survey, out of

1000 participating officers, 84% felt that stress and fatigue were widely prevalent

aboard ships, (Cole-Davies 2001).

In addition, Dr. Rothblum [80] indicates that the NTSB has identified fatigue to be an

important cross-modal issue, being just as pertinent and in need of improvement in the

maritime industry as it is in the aviation, rail, and automotive industries. Several

studies, some of which were mentioned earlier, have cited fatigue as the primary

factor for mariner‘s underperformance.

In conclusion, the importance of fatigue lays on the fact that even communications or

incompetency errors can be the result of a fatigued individual. For example, a sleep

deprived operator can make a wrong judgment in an emergency situation. The

probability of this judgment error would be significantly lower if the operator had

slept enough since fatigue affects directly an individual‘s cognitive skills. However,

this error would be recorded as decision issue rather than a fatigue related one (due to

the underreporting of fatigue) thus creating a trend to focus on the enhancement of

decision making culture rather than mitigation of fatigue aboard.



CHAPTER 3 FATIGUE



3. FATIGUE

3.1. WHAT IS FATIGUE IN TERMS OF PHYSIOLOGY

Fatigue can be described as the feeling of tiredness or weakness that reduces the

ability to perform ordinary tasks. Fatigue affects everyone differently. It can result

into an intensive feel of tiredness resulting in a great urge for sleep. Pain can

sometimes come along with fatigue as well. In general, it is the human body's way of

signaling its need for rest and sleep. If there is a persistent feeling of tiredness or

exhaustion present, which goes beyond normal sleepiness, it is usually the ‗alarm‘ that

something serious is amiss.

Physically, fatigue is characterized by a profound lack of energy, feelings of muscle

weakness, and slowed movements or central nervous system reactions. Fatigue can

also trigger serious mental exhaustion. Persistent fatigue can cause a lack of mental

clarity, difficulty concentrating, and in some cases, even lack of memory. Chronic

fatigue syndrome (CFS) is a long-term fatigue that may occur with other symptoms,

such as recurrent sore throats, muscle pain, multi-joint pain, tender lymph nodes, new

patterns of headaches, and complaints such as impaired memory or concentration

[BB].

According to handicapsupport.com (L) fatigue can be dangerous when the tasks to be

performed require great deal of concentration. When an individual is sufficiently

fatigued, he or she may fall asleep for a small period while on duty, a common fatigue

effect known as microsleep. However, objective cognitive testing should be done to

pinpoint the difference between the neurocognitive deficits of brain disease from

those attributed to fatigue.

Fatigue can be found in literature by various aliases such as

Exhaustion

Lethargy

Languidness

Languor

Lassitude or Littleness

Fatigue is divided in two subcategories and two types

1. Physical fatigue

2. Mental fatigue

And

A) Central fatigue

B) Peripheral fatigue



Smith et al. [90] indicate that the technical use of the term fatigue is imprecise. The

variety of fatigue inducing situations, time courses and outcomes suggests that it is

unlikely that we are considering a single set of processes leading to a specific

underlying state. An individual that is affected by fatigue will experience possible

performance deterioration and a probable physiological malfunctioning. Therefore,

acute fatigue can be recognized as the main source of the symptoms that come with it.

Reporting of fatigue and presence of symptoms is the first step in assessing the issue.

Moreover, Smith pinpoints that fatigue is heavily dependent on which aspect of the

fatigue process one uses as the indicator of its presence. For example, if one assumes

that doing shift work is a risk factor for fatigue one might simply use the number of

workers doing shift work as an indicator of prevalence. However, this is based on the

assumption that shift work automatically leads to fatigue which one finds is not

always the case. Similarly, fatigue may be measured by the presence of negative

outcomes, but the extent of the problem will often depend on the indicator chosen.

There is no single ―right‖ approach: all aspects of the fatigue process must be assessed

and considered [90].

The main question that remains is: Why is fatigue so important?

In their recent research regarding Human Factors, HSE [C] indicated that poorly

designed shift-working arrangements and long working hours that do not balance the

demands of work with time for rest and recovery can result in fatigue, accidents,

injuries and ill health.

Consequently, fatigue can often be synonymous to excessive working time, poorly

designed shift patterns and continuous lack of sufficient rest. It may come more easily

if the tasks to be performed are complex, tedious and machined paced so it is deeply

connected to the nature of the workload.

Fatigue has played an integral part in numerous accidents, ill-health and injury, and

reduced productivity. Major disasters or near misses such as Herald of Free

Enterprise, Chernobyl, Texas City, Clapham Junction, Challenger and Exxon Valdez

have all been related to fatigue presence and outcomes.

HSE statistical data show that Fatigue has been implicated in 20% of accidents on

major roads and is said to cost the UK £115 - £240 million per year in terms of work

accidents alone. These costs might be an overestimation but if we take fatigue

underreporting into consideration, the actual monetary damage could be even greater.

For example, the restoration of a large oil spill (that can easily be the result of fatigue

in the maritime domain) may cost more than a billion U.S. $.



Figure 3. 1: The Chernobyl nuclear accident

3.1.1 PHYSICAL FATIGUE

Physical fatigue or muscle weakness is a direct term for the inability to exert force

with one‘s muscles to the degree that would be expected given the individual‘s

general physical fitness. A test of strength is often used during a diagnosis of a

muscular disorder before the actual etiology can be identified. Such etiology depends

on the type of muscle weakness, which can be true or perceived as well as central or

peripheral. True weakness is substantial, while perceived rather is a sensation of

having to put more effort to do the same task. On the other hand, central muscle

weakness is an overall exhaustion of the whole body, while peripheral weakness is an

exhaustion of individual muscles [M].

3.1.2 MENTAL FATIGUE

In addition to physical, fatigue also includes mental fatigue, not necessarily including

any muscle fatigue. Mental fatigue describes the effect that people experience during

and following prolonged periods of cognitive activity that requires work efficiency.

Such a mental fatigue, in turn, can manifest itself either as decreased wakefulness or

just as a general lack of attention, not necessarily including sleepiness. It may also be

described as a more or less decreased level of consciousness. In any case, this can be

dangerous when performing tasks that require constant concentration, such as driving

a vehicle or paroling.

To summarize, mental fatigue refers to the state caused by extended period of

demanding cognitive activity. The effect of mental fatigue on cognitive skills and

http://en.wikipedia.org/wiki/Muscle

http://en.wikipedia.org/wiki/Physical_fitness

http://en.wikipedia.org/wiki/Medical_diagnosis

http://en.wikipedia.org/wiki/Etiology

http://en.wikipedia.org/wiki/Wakefulness

http://en.wikipedia.org/wiki/Attention

http://en.wikipedia.org/wiki/Level_of_consciousness

http://en.wikipedia.org/wiki/Vehicle



mental performance are rather profound. However, the impact of mental fatigue on

physical performance is often neglected and not sufficiently researched.

An interesting comparison between physiological and psychological fatigue is made

by Shen et al. [86]. In this essay, physiological fatigue is described as a loss of

maximal force generating capacity during muscular activity or a failure of the

functional organ. It may be induced by excessive energy consumption or the depletion

of hormones, neurotransmitters or essential physiological substrates. Physiological

fatigue may be associated with fever, infection, anemia, sleep disturbances, and

pregnancy. Psychological fatigue, in contrast, has been defined as a state of weariness

related to lack of motivation. Psychological fatigue has been associated with stress

and other intense emotional experiences and may accompany depression and anxiety

[86].

3.1.3 CENTRAL FATIGUE

The central component to fatigue is generally described in terms of a reduction in the

neural drive or nerve-based motor command to working muscles that result in a

decline in the force output. It has been suggested that the reduced neural drive during

exercise may be a protective mechanism to prevent organ failure if the work was

continued at the same intensity [M].

3.1.4. PERIPHERAL FATIGUE

Fatigue during physical work is considered an inability for the body to supply

sufficient energy to the contracting muscles to meet the increased energy demand.

This is the most common case of physical fatigue-affecting a national average of 72%

of adults in the work force in 2002. This causes contractile dysfunction that is

manifested in the eventual reduction or lack of ability of a single muscle or local

group of muscles to do work. The insufficiency of energy, i.e. sub-optimal aerobic

metabolism, generally results in the accumulation of lactic acid and other acidic

anaerobic metabolic by-products in the muscle, causing the stereotypical burning

sensation of local muscle fatigue.

The fundamental difference between the peripheral and central theories of fatigue is

that the peripheral model of fatigue assumes failure at one or more sites in the chain

that initiates muscle contraction. Peripheral regulation is therefore dependent on the

localized metabolic chemical conditions of the local muscle affected, whereas the

central model of fatigue is an integrated mechanism that works to preserve the

integrity of the system by initiating fatigue through muscle derecruitment, based on

collective feedback from the periphery, before cellular or organ failure occurs.

Therefore the feedback that is read by this central regulator could include chemical

and mechanical as well as cognitive cues. The significance of each of these factors

will depend on the nature of the fatigue-inducing work that is being performed [CC].



The following image shows a typical example of central (left) and peripheral fatigue

(right). Peripheral fatigue, as shown in figure 3.2, can be the result of a specific type

of exercise such as working out the biceps. The red zone represents the impaired part

of the body that needs to rest, in order to perform the same task effectively.

Figure 3. 2: Central and Peripheral fatigue

In a similar comparison to that between physiological and psychological fatigue Shen,

Barbera et al [90] state that central models of fatigue imply a malfunction of the CNS

(central nervous system), such as impaired transmission between the CNS and the

peripheral nervous system, or dysfunction of selected areas of CNS such as the

hypothalamic region. Peripheral models of fatigue, in contrast, view fatigue as

resulting from dysfunction of the peripheral nervous system, such as impaired

neuromuscular transmission at the motored-plate. While dualistic approaches have

proven to be popular, such definitions fail to capture the multidimensional nature of

the fatigue. A number of other definitions for fatigue have been proposed [90].

3.2. FATIGUE IN MEDICAL TERMS

In medical physiology, fatigue is the inability to perform reasonable and necessary

physical or mental activity. When the metabolic reserves of the body are exhausted

and the waste products increased, as for example after prolonged exertion, the body

finds it difficult to continue its function and activity. The accumulation of lactic acid

http://www.questia.com/library/encyclopedia/101254144



in muscle tissue and the depletion of glycogen (stored glucose) results in muscle

fatigue. The contractile properties of muscle are reduced, and continued exertion is

impossible unless the muscle is allowed to rest. In the normal body a period of rest

permits redistribution of nutritive elements to the muscles and tissues and elimination

of accumulated waste products; the body is then ready to resume activity. There are

some persons in whom fatigue is a chronic state that does not necessarily result from

activity or exertion. In some instances this abnormal fatigue may be associated with

systemic disorders such as anemia, a deficiency of protein or oxygen in the blood,

addiction to drugs, increased or decreased function of the endocrine glands, or kidney

disease in which there is a large accumulation of waste products. If excessive fatigue

occurs over a prolonged period, exhaustion (marked loss of vital and nervous power)

may result. In most persons with chronic fatigue, however, the condition seems to be

associated with manic-depressive disorder. Thorough medical and psychiatric

examination may be required [N].

Figure 3. 3: Muscle impairment and rejuvenation process

3.3 FATIGUE RISK FACTORS

In their research regarding seafarer‘s fatigue Smith et al. [94] summarize t that fatigue

may be the result of a great number of factors: lack of or poor quality sleep, long

working hours, working at times of low alertness, prolonged work, insufficient rest

between work periods, excessive workload, noise and vibration, motion, medical

conditions and illnesses. Chronic fatigue can either be the result of continuous

exposure to acute fatigue or can represent a failure of rest and recuperation to remove

http://www.questia.com/library/encyclopedia/101246524



fatigue. Many working patterns induce acute fatigues which in time lead to chronic

fatigue. For example, working at night is associated with reduced alertness during the

shift and may also produce cumulative problems because of poor sleep during the day.

Risk factors for fatigue have been deeply connected with problems related to sleep

quality and deprivation as well as day-time sleep following a night shift.

In addition, Smith et al. [94] state that risk factors for fatigue have been categorized

by researchers into factors that reflect the work environment (e.g. working hours, task

nature, the physical surroundings) and aspects related to the individual (both stable

traits, and current state). It is of the outmost importance to recognize that it is the

interplay of these risk factors that is significant; fatigue is most likely to appear when

a large number of these factors are present. It is a common fact that most regulatory

bodies have, until recently, focused their research and legislation regarding fatigue on

work schedules. On the other hand, the role of psychological and emotional factors is

rather neglected resulting in insufficient research on their sources and impact.

Moreover, Smith et al. recognizes that only few studies have examined how risk

factors might interact in terms of their effects, or attempted to pinpoint the different

risk factors (e.g. what are the relative contributions of factors such as sleep

deprivation, long working hours and high job demands to fatigue levels?). Recent

studies regarding fatigue have indicated that psychosocial workplace stressors tend to

demonstrate cumulative associations with self reports of work stress and poor health

outcomes with similar or greater severity to those caused by physical fatigue [94].

A general summary of the most common and influential fatigue risk factors is

presented by Calhoun [14]:

poor sleep quality

sleep deprivation

physical/mental exertion

emotional stress

disruption of circadian rhythms

poor physical condition,

And drug/alcohol use.

3.3.1 SLEEP & SLEEP QUALITY

According to Macmillan Dictionary (1980) sleep is a naturally recurring altered state

of consciousness with relatively suspended sensory and motor activity, characterized

by the inactivity of nearly all voluntary muscles. It is distinguished from quiet

wakefulness by a decreased ability to react to stimuli, but it is more easily reversible

than hibernation or coma. Sleep is a heightened anabolic state, accentuating the

growth and rejuvenation of the immune, nervous, skeletal and muscular systems [O].

http://en.wikipedia.org/wiki/Altered_state_of_consciousness

http://en.wikipedia.org/wiki/Altered_state_of_consciousness

http://en.wikipedia.org/wiki/Voluntary_muscle

http://en.wikipedia.org/wiki/Wakefulness

http://en.wikipedia.org/wiki/Stimulus_%28physiology%29

http://en.wikipedia.org/wiki/Hibernation

http://en.wikipedia.org/wiki/Coma

http://en.wikipedia.org/wiki/Anabolic



The need for sleep and its physiology have being researched extensively for many

years. Research has shown that the human body needs "restorative sleep" in order to

remain alert and rejuvenate its „batteries‟. In order to have restorative sleep four

stages of sleep are necessary to take place. Table 3.1 shows four general factors that

constitute the overall picture of sleep quality.

Table 3. 1: Factors that affect the overall sleep quality [14]

3.3.1.1. SLEEP DURATION (QUANTITY)

Restorative sleep plays a significant role on individual performance so for avoiding

sleep deprivation it is essential for the human body to establish a sleep quantity

routine. Individuals have different requirements regarding the hours of sleep needed

to feel fresh and renewed the following morning. The vast majority of the human

population requires seven to eight hours. Alertness levels in correlation to sleep

duration are shown below in Figure 3.4 [14].We can see that the turning point from

reduced to full alertness lies within four to six hours of sleep. However, this figure

refers to person who had regular sleeping patterns in the past. If an individual sleeps

an average of five hours for a week, he is likely to experience lack of alertness due to

sleep deprivation.



Figure 3. 4: The effect of sleep deprivation on alertness,(Sirois, 1998.)1[14]

It is a common fact that if a person sleeps less than five hours, there is a great

probability to be drowsy when he gets up the next day. Figure 3.4 also shows that the

average person needs eight or more hours of sleep in order to reach peak levels of

alertness [14].

Medical research has shown that the sleeping time each person needs depend on

various factors, the most important being age. In figure 3.5 we can see an empirical

estimation of the sleeping hours needed per day for a number of different age groups.

While these needs are representative, each individual might require more or less sleep

than the average due to general health, environmental and social issues. For example,

research has shown that an individual with a regular smoking habit might need two

extra hours of sleep to that of a non-smoking person of the same age.

1 In the above graph, the measure of alertness is the multiple sleep latency test

(MSLT). This is a measure of how quickly a person falls asleep. An alert person takes

the most time and a fatigued person takes less time



Figure 3. 5: Average hours of sleep needed per day-Age [O]

3.3.1.2 SLEEP ARCHITECTURE

Sleeping quantity is an essential part of the whole sleep quality procedure. While

asleep, the human body cycles through different levels of sleep. These levels are

known as sleep stages and each serves a different purpose for rejuvenating the body

(the purpose of each stage is currently researched). This cyclical pattern of light sleep,

deep sleep and rapid eye movement (REM) sleep is called sleep architecture. In the

following table we can see the duration, effects and order of all 4 known sleep stages.

Table 3. 2: Sleep architecture stages [14]



The entire cycle normally occurs in 90-minute periods, providing approximately four

to five cycles in an 8-hour period. The proportions and duration of each stage are

critical to the restorative quality of the cycle. Research has shown that restorative

sleep requires three to five uninterrupted sleep cycles.

Experts currently believe that the third and fourth sleep stages are responsible for

body restoration and REM for strengthening and organizing memory. The cycle

continues throughout the sleep period, with each REM stage increasing in length.

Interruptions while sleeping, such as those caused by light, noise, and excessive

vibration, tend to keep the body in the lighter stages. This makes it difficult to

transition to deeper sleep, thereby reducing the ability to get the important third,

fourth, and fifth stages that are required for sleep to be restorative. It is very probable

that even one day of reduced sleep is enough for a person to experience a decline in

their alertness levels. The loss of sleep over a week‘s time causes dangerously low

levels of alertness and this is when accidents are likely to occur. These same effects

are experienced by those who work late at night and through the early morning hours

[14].

3.3.1.3 COMMON CAUSES OF SLEEP LOSS

The Queensland government website [H] indicates that a number of factors in the

workplace and in a person's private life can cause sleep loss. Examples from the

workplace include:

Extended working hours

Irregular and unpredictable working hours

Time of day when work is performed and sleep obtained

Shift work

Having more than one job and

Stress.

3.3.1.4 SLEEPING ENVIRONMENT

Sleeping environment is an integral part of the overall sleep quality. The idea of a

satisfying sleep environment rests on a plethora of factors. The most important factor

is the temperature. The brain acts slowly if its temperature is in a comfortable zone.

Temperature controls the brain activity and in turn; the body activity. If it is cool it

affects the brain to be in rest and one can sleep. Another factor is light - we need

darkness or toned light - the mixture of dark and light. The third factor is sound.

When we close our eyes we feel comfortable, but if sound is there our eyes will open

so we need a quiet environment. The other factor is the surface of the mattress. The

smoother the surface is, the better we will sleep. To summarize, the key contributors

for a healthy sleeping environment are:



Light

Noise

Temperature

Some other factors that have been known to worsen the overall sleep

environment are:

Humidity

Vibrations

Uncomfortable clothing

Pillows and mattresses

Smoky environment

Available space

The following figure (3.6) shows the difference between a healthy and a hazardous

environment. It should be mentioned that in hazardous sleeping environment the

individual is often vulnerable to the factors that can disrupt sleep continuity and

decrease sleep quality.

Figure 3. 6: Examples of hazardous (left) and healthy (right) sleeping environments

3.3.1.5 TIME OF DAY

Many scientists have indicated that the time of the day can also be a contributing

factor to the sleep quality. It is a common fact that the human body is not designed to

operate at night and sleep during the day, so it is rather clear that sleeping during

daytime could prove extremely difficult. A graphical depiction of why this is true is

shown in Figure 3.7 (Sirois, 1998). The graph shows changes in body core



temperature throughout the day, which directly relates to alertness. A drop in

temperature naturally causes sleepiness.

Figure 3. 7: Body temperature during a regular day (Sirois, 1998) [103]

3.3.1.6 NAPPING

A nap is a short period of sleep, usually in the daytime. Naps may be taken when one

person becomes drowsy during the day or as a traditional daily practice. It is common

for small children and elderly people to take frequent naps. Experts agree that the best

way to fight fatigue is to get enough sleep every night. But for some people,

especially those who work long hours, have care giving responsibilities or work at

night, this can be an ambitious goal. Even people who do get enough sleep regularly

may feel the effects of the midday dip, especially after a heavy meal. Studies show

that taking a nap is a great way to increase alertness and reaction times, improve

mood, and reduce accidents. For many people, napping is also a highly pleasurable

experience [P].

3.3.1.7 SLEEP DEPRIVATION

Sleep deprivation, a sleep disorder characterized by having too little sleep, can be

either chronic or acute. Long-term sleep deprivation causes death to lab animals. A

chronic sleep-restricted state can cause fatigue, daytime sleepiness, clumsiness and

weight loss or weight gain, [O]. Extended periods between sleep sessions deprive a

person of the rest needed to remain alert. Sleep deprivation drastically reduces

alertness levels after a person has been awake for more than fourteen hours. Figure

3.8, (Sirois, 1998) [14] illustrates this point:

http://en.wikipedia.org/wiki/Drowsy

http://en.wikipedia.org/wiki/Sleep_disorder

http://en.wikipedia.org/wiki/Sleep



Figure 3. 8: Sleep deprivation and alertness 14]

Common effects of sleep deprivation

The end-your-sleep-deprivation website [Z] categorizes the effects of sleep

loss/deprivation in the following:

1. Decreased Alertness and Ability to Maintain Focus

Carrying a sizable sleep debt throughout the day can drastically decrease

productivity. Fatigue will compromise your attention, and as a result cognitive

performance will suffer. Specifically, learning, memory, and creativity are

frequently hampered by a large sleep debt. In a situation such as driving a car

this decreased alertness can, and has repeatedly, led to fatal results.

2. Mood

It's not uncommon for sleep-deprived individuals to be subject to extreme

emotions and mood swings. A very tired person who is laughing

uncontrollably at one moment may be crying or yelling angrily a few minutes

later. In addition, sleep deprived individuals tend to be short-tempered and

difficult to communicate with while fatigue also results in lack of motivation.

3. Energy and Motivation

A decrease in energy and motivation is probably the most noticeable

consequence of sleep deprivation. Individuals who have not received sufficient

sleep will feel lethargic and uninspired to work.



4. Control, Coordination, and Impulsiveness

Lack of sleep is often associated with a hindrance of bodily control. Tired

individuals often feel enhanced physical impulses, such as an otherwise

inexplicable desire to eat.

3.3.2 CIRCADIAN RHYTHMS

According to Dr. Samuel Strauss, "Circadian rhythms" are physiological and

behavioral processes, such as sleep/wake, digestion, hormone secretion, and activity

that oscillate on a 25 hour basis. Each rhythm has a peak and a low point during every

day/night cycle. Time cues, called "zeitgebers" keep the circadian "clock" set to the

appropriate time of day. Common zeitgebers include daylight, meals and work/rest

schedules. If the circadian clock is moved to a different schedule, for example when

crossing time zones or changing from a day work shift to a night shift, the resulting

"sleep phase shift" requires a certain amount of time to adjust to the new schedule.

This amount of time depends on the number of hours the schedule is shifted, and the

direction of the shift. During this transition, the circadian rhythm disruption or "jet

lag" can produce effects similar to those of sleep loss [Q].

Historically, to differentiate genuinely endogenous circadian rhythms from

coincidental or apparent ones, three general criteria must be met: 1) the rhythms

persist in the absence of cues, 2) they persist equally precisely over a range of

temperatures and 3) the rhythms can be adjusted to match the local time [DD]:

The rhythm persists in constant conditions (for example, constant dark) with a

period of about 24 hours. The rationale for this criterion is to distinguish

circadian rhythms from those "apparent" rhythms that are merely responses to

external periodic cues. A rhythm cannot be declared to be endogenous unless

it has been tested in conditions without external periodic input.

The rhythm is temperature-compensated, i.e., it maintains the same period

over a range of temperatures. The rationale for this criterion is to distinguish

circadian rhythms from other biological rhythms arising due to the circular

nature of a reaction pathway. At a low enough or high enough temperature, the

period of a circular reaction may reach 24 hours, but it will be merely

coincidental.

The rhythm can be reset by exposure to an external stimulus. The rationale for

this criterion is to distinguish circadian rhythms from other imaginable

endogenous 24-hour rhythms that are immune to resetting by external cues

and, hence, do not serve the purpose of estimating the local time. Travel across

time zones illustrates the necessity of the ability to adjust the biological clock

so that it can reflect the local time and anticipate what will happen next. Until

rhythms are reset, a person usually experiences jet lag.

http://en.wikipedia.org/wiki/Time_zone

http://en.wikipedia.org/wiki/Jet_lag



Figure 3.9 gives us a rough representation of the human ―biological clock‖ of an

average individual [DD].

Figure 3. 9: Biological clock and Circadian Rhythms [DD]

This natural rhythm of alertness is illustrated below in Figure 3.10 where we can

easily pinpoint the peak and the drowsy periods of the circadian rhythms.

Figure 3. 10: Circadian Rhythms and Alertness [14]



The biological clock regulates energy cycles so that alertness increases after wake-up

time, peaks in the mid-morning hours, dips through the late morning and early-

afternoon hours, peaks again in the early-evening hours, and then decreases

throughout the late-evening and early-night hours, reaching a daily low in the middle

of the night. The exact times of these peaks and valleys depend on specific inputs to

the biological clock system, namely wake-up times, bedtimes, and daily interval of

daylight (or artificial bright light) exposure [103].

The circadian rhythms most dominant attribute is its effect on the human sleep cycle;

this combined with its influence on body temperature regulation, hormone production

and adrenal gland output, cause the body to desire sleep. Sleep is desired and initiated

at preferred times relative to the circadian rhythm of core body temperature (Monk,

1987). In general, the longest and best quality sleep episodes are initiated several

hours prior to the body temperature minimum. Humans are most productive when

their body temperature is at its highest and logically at its least efficient, when

temperature is lowest. Human performance degradation at circadian lows is one of the

major challenges for the aviation industry [A].

The NASA Ames Fatigue Countermeasures Program has underlined the fact that

sleep loss, and circadian disruption lead to significant decrements in alertness and

performance, (Neri et al., 2002). A 1995 Airbus report highlights Human performance

decrements include problems with:

Vigilance

Alertness

Irritability

Loss of mental agility

Ability to multi-task

Decision making

Perceptive skills

The challenge facing flight crews is the need to maintain vigilance during long,

highly-automated, and often boring night flights. Traditionally most accidents and

incidents occur during the approach and landing phase of flight [A].

The most common causes of circadian disruption are:

Jet Lag: The need to adjust ones biological clock to global time changes in a

short period. This issue is very common in aviation and truck drivers.

Routine change or shift change: When the daily sleep routine is altered due to

various circumstances. Airline pilots and seafarers often ‗swallow‘ shift

changes which do not allow them to establish a healthy daily sleep routine

thus disrupting their ‗biological clock‘.



To conclude, circadian rhythms are responsible for coordinating the following

physiological aspects.

Sleepiness/wakefulness

Body temperature

Digestion

Cardiovascular responses

Digestion

Alertness/Vigilance

3.3.3 SECONDARY RISK FACTORS

Apart from sleep and circadian rhythms, there are numerous other fatigue risk factors.

It is rather clear that their impact is of lesser impact of the two factors discussed

above. However, these factors are not to be ignored since their effects and interaction

with each other favor high levels of fatigue. Moreover, most of them have interplay

with both sleep quality and the disruption of circadian rhythms, increasing indirectly

fatigue presence in work and domestic environments.

Some of the most common secondary fatigue risk factors are discussed in the

following paragraphs.

3.3.3.1 PHYSICAL CONDITION

The level of fatigue increases as physical condition drops. Fitness, which is directly

associated with physical condition can be an important factor in reducing the presence

and effects of fatigue.

Figure 3.11 shows the work done by a hotspot crew in a 9-day period. The crew was

divided in fitter and less fit groups. The fitter group did more work per day. It can be

inferred that the effect of fatigue was greater on the less fitter group thus effecting

their performance.

Aspects that have impact on physical condition are:

1. Exercise

2. Sufficient and healthy nutrition

3. Illness

4. Alcohol

5. Medication(per scripted)

6. Substances (drugs etc.)

7. Smoking habits

8. Sleep disorders



Figure 3. 11: Fatigue-fitness connection [105]

3.3.3.2 ENVIRONMENT

Fatigue is deeply connected to the environmental surroundings. A hazardous and non-

ergonomic environment can decrease the performance and endanger the safety of an

individual. A very common example of this assumption is the idea of performing a

certain task in the presence of extreme heat. Heat has a severe impact on body

temperature causing an outburst of fatigue. Fatigue factors that are linked to the

environment are:

1. Temperature

2. Noise

3. Weather phenomena

4. Humidity

5. Lighting

6. Vibrations (moving environments aircrafts, ships etc.)

7. Ventilation

3.3.3.3 WORK WONDITION

Recent research has shown that job demands and overall nature play a key role in the

emergence of fatigue. Statistical data and scientific expertise agree that job-specific

fatigue is affected by:

1. Workload: The amount of workload is logically connected to fatigue.

Heavy workload often means limited time for rest and an inevitable

concentration of fatigue. Moreover, workload, by means of shift



change, can possibly disrupt the circadian rhythms as mentioned in

paragraph 3.3.2.

2. Type of work: Type of work in terms of job ‗attractiveness‘ can often

result in fatigue problems. A tedious and repetitive task can provoke

tiredness and even sleepiness. On the other hand, a pleasant and

interesting activity could reduce potential signs of fatigue, especially

mental.

3.3.3.4 STRESS

According to Selye (1956) stress is a term in psychology and biology, first coined in

the biological context in the 1930s, which has in more recent decades become a

commonplace of popular parlance. It refers to the consequence of the failure of an

organism – human or animal – to respond appropriately to emotional or physical

threats, whether actual or imagined.

Stress symptoms commonly include a state of alarm and adrenaline production, short-

term resistance as a coping mechanism, and exhaustion, as well as irritability,

muscular tension, inability to concentrate and a variety of physiological reactions such

as headache and elevated heart rate [T].

The Université Laval occupational health and safety management department [U]

states that the transactional model of Lazarus defines stress as an imbalance between

the demands of the environment and the individual's resources. According to this

model, the individual makes a primary appraisal of the situation or the demand with

which he is confronted. The demand may be perceived as a challenge or even a threat

if the individual thinks that it may give rise to negative consequences-and-extent. The

individual then makes a secondary appraisal where he attempts to determine what

resources are available to meet the demand. Thus, workplace stress implies that work-

related demands exceed the employee's ability to adapt to these demands. The effects

of stress may therefore be positive and provide the motivation, energy and creativity

needed to accomplish a task if the individual thinks that he has the abilities and

resources needed to succeed at it. The effects of stress will be negative when there is a

discrepancy between the individual's resources and the demand.

Here are many causes (stressors) of stress in life including:

Death: of spouse, family, friend

Health: injury, illness, pregnancy

Crime: Sexual molestation, mugging, burglary, pick-pocketed

Self-abuse: drug abuse, alcoholism, self-harm

Family change: separation, divorce, new baby, marriage

Sexual problems: getting partner, with partner

Argument: with spouse, family, friends, co-workers, boss

http://en.wikipedia.org/wiki/Psychology

http://en.wikipedia.org/wiki/Biology

http://en.wikipedia.org/wiki/Vernacular

http://en.wikipedia.org/wiki/Organism

http://en.wikipedia.org/wiki/Human

http://en.wikipedia.org/wiki/Animal

http://en.wikipedia.org/wiki/Emotion

http://en.wikipedia.org/wiki/Body

http://en.wikipedia.org/wiki/Symptom

http://en.wikipedia.org/wiki/Fight-or-flight_response

http://en.wikipedia.org/wiki/Adrenaline

http://en.wikipedia.org/wiki/Exhaustion

http://en.wikipedia.org/wiki/Irritability

http://en.wikipedia.org/wiki/Attention

http://en.wikipedia.org/wiki/Physiology

http://en.wikipedia.org/wiki/Headache

http://en.wikipedia.org/wiki/Heart_rate



Physical changes: lack of sleep, new work hours

New location: vacation, moving house

Money: lack of it, owing it, investing it

Environment change: in school, job, house, town, jail

Responsibility increase: new dependent, new job.

UK's HSE lists six key stress factors (stressors) at work [C]:

1. The demands of the job

2. The control staff have over how they do their work

3. The support they receive from colleagues and superiors

4. Their relationships with colleagues

5. Whether they understand their roles and responsibilities

6. How far the company consults staff over workplace changes.

In addition, Universite Laval website [U] states some common symptoms of stress:

Physical symptoms

Cardio-vascular disorders

Allergies

Dermatological disorders

Migraines

Respiratory disorders

Sleep disorders

Gastrointestinal disorders

Psychological symptoms

Depression

Anxiety

Boredom

Frustration/Irritability

Isolation

Difficulties concentrating or making decisions

Memory lapses

Behavioural symptoms

Aggressivity

Alcohol or drug abuse

Eating disorders

Conflicts

Absenteeism

Decreased productivity



Decision to leave job

Accident proneness

The Consequences of Stress for the individual

It is rather obvious that stress affects fatigue in many different ways. The direct

consequences of stress can be summarized in the following categories.

1. Mood disorders

2. Anxiety disorders

3. Burn out (unhealthy relationship with work environment)

4. Survivor syndrome

3.4. EFFECTS OF FATIGUE

Cumulative fatigue can have numerous various effects in everyday life. In this

paragraph the most considerable results of fatigue will be presented. The complex

nature of fatigue and its causes makes the number of consequences too great to count.

However, scholars and researchers have categorized these effects in three greater

categories

3.4.1. WORK PERFORMANCE

3.4.1.1 GENERAL

High levels of fatigue cause reduced performance and productivity, and increases the

risk of accidents and injuries. Fatigue affects the ability to think clearly. As a result

people who are fatigued are unable to gauge their own level of impairment, and are

unaware that they are not functioning as well or as safely as they would be if they

were not fatigued. Performance levels drop as work periods become longer and sleep

loss increases. Staying awake for 17 hours has the same effect on performance as

having a blood alcohol content of 0.05%. Staying awake for 21 hours is equivalent to

a blood alcohol content of 0.1% [H].

The most common effects associated with fatigue are:

Desire to sleep

Lack of concentration

Impaired recollection of timing and events

Irritability

Poor judgement

Reduced capacity for communicating with others

Reduced hand-eye coordination

Reduced visual perception

Reduced vigilance



Reduced capacity to judge risk and

Slower reaction times.

Not only do these effects decrease performance and productivity within the

workplace, but they simultaneously increase the potential for incidents and injuries to

occur. People working in a fatigued state may place themselves and others at risk,

most particularly:

When operating machinery (including driving vehicles)

When performing critical tasks that require a high level of concentration and

Where the consequence of error is serious.

3.4.1.2 SLEEP INERTIA

Sleep inertia is that groggy feeling that is experienced for up to a half-hour after

waking up. Managers and safety observers must be aware of sleep inertia and should

act to prevent someone from performing tasks immediately after waking up. This is a

very common problem because it always affects someone who just woke up, whether

fatigued or not. Waking up from a deep sleep simply requires time in order to become

oriented and to raise awareness [14].

3.4.2 MICRO SLEEPS

A micro sleep is a brief nap that lasts for approximately four to five seconds. People

who suffer from micro sleeps are not always aware when a micro-sleep occurs, which

can have a significant effect on safety. Micro sleep is the most dangerous and scary

effect because the person actually falls asleep for short periods, as long as ten to

fifteen seconds. When performing a dangerous task or monitoring safety critical

evolutions, falling asleep for any amount of time can be fatal. Micro sleep is what

often causes car accidents and results in fatalities on dangerous job sites.

Figure 3.12 shows a driver experiencing a microsleep on the wheel. Fatigue resulting

to microsleeps has been a common issue in the maritime industry as well. Many

groundings have occurred because of a fatigued watchkeeper who experienced a

microsleep or even fell asleep in duty. Moreover, the consequences of microsleep

while driving can be fatal. Due to the nature of this activity, the time available to react

in such situations is limited. Many tired drivers have found themselves driving in the

wrong side of the road or even wake up when their car crashed on an obstacle. Several

surveys have pinpointed that accidents related to microsleeps constitute a considerable

percentage of all occurring road incidents.



Figure 3. 12: Driver experiencing a microsleep

3.4.3 HEALTH EFFECTS

3.4.3.1 GENERAL HEALTH ISSUES

According to the Queensland Government website [H], the effects of fatigue increase

with age. People over 50 years of age tend to have lighter, fragmented sleep; which

can prevent them from receiving the recuperative effects from a full night of sleep,

and can make them more likely to become fatigued.

Lack of sleep has been indirectly linked with the following health effects:

Heart disease and high blood pressure

Stomach disorders

Mental illnesses and

Lower fertility.

When the circadian rhythm is disrupted, the treatment of some medical conditions can

be affected. Examples of medical conditions which may be affected include:

Asthma

Depression and

Diabetes.

http://www.deir.qld.gov.au/workplace/subjects/fatigue/effects/health/index.htm#heart

http://www.deir.qld.gov.au/workplace/subjects/fatigue/effects/health/index.htm#stomach

http://www.deir.qld.gov.au/workplace/subjects/fatigue/effects/health/index.htm#mental

http://www.deir.qld.gov.au/workplace/subjects/fatigue/effects/health/index.htm#fertility



3.4.3.2 CHRONIC FATIGUE SYNDROME

Chronic fatigue syndrome (CFS) is the most common name given to a variably

debilitating disorder or disorders generally defined by persistent fatigue unrelated to

exertion, not substantially relieved by rest and accompanied by the presence of other

specific symptoms for a minimum of six months. The disorder may also be referred to

as post-viral fatigue syndrome (PVFS), when the condition arises following a flu-like

illness, myalgic encephalomyelitis (ME), or several other terms. The disease process

in CFS displays a range of neurological, immunological, and endocrine system

abnormalities. Although classified by the World Health Organization under Diseases

of the nervous system, the etiology (cause or origin) of CFS is currently unknown and

there is no diagnostic laboratory test or biomarker [M].

Common symptoms of fatigue include:

Impaired memory or concentration

Post-exertional malaise, where physical or mental exertions bring on "extreme,

prolonged exhaustion and sickness"

Unrefreshing sleep

Muscle pain (myalgia)

Pain in multiple joints (arthralgia)

Headaches of a new kind or greater severity

Sore throat, frequent or recurring

Tender lymph nodes (cervical or axillary)

3.5 FATIGUE-RELATED ACCIDENTS

Caldwell (2001) [12] makes a historical perspective on select fatigue-related problems

in the industrial sector. A typical example of this was in 4 AM on March 28, 1979,

Unit No. 2 at the Three Mile Island Nuclear Power Plant shut down as nearly every

valve in the feed water cooling system slammed shut, a pump casing ruptured, and

numerous pipes shattered. Given the lack of a cooling system, both heat and pressure

in the reactor began to build to dangerous levels, and a plant operator‘s failure to

recognize that emergency backup pumps were not functioning began a chain of events

that nearly resulted in a nuclear meltdown. Two days of intense effort on the part of

plant workers and Nuclear Regulatory Commission personnel ultimately brought the

situation under control but not without significant costs. The consequences were that

radioactive gases were vented into the air, radioactive water was released into the

Susquehanna River, and a reactor whose 36,000 fuel rods ruptured after reaching

temperatures of 4300 degrees was lost. Thankfully, a major nuclear catastrophe was

avoided.

Another well-known example came seven years later, when the former Soviet Union

was not so fortunate. In the early morning hours of April 26, 1986, the sequence of

http://en.wikipedia.org/wiki/Post-viral_fatigue_syndrome

http://en.wikipedia.org/wiki/Alternative_names_for_chronic_fatigue_syndrome

http://en.wikipedia.org/wiki/Etiology

http://en.wikipedia.org/wiki/Memory

http://en.wikipedia.org/wiki/Malaise

http://en.wikipedia.org/wiki/Pain

http://en.wikipedia.org/wiki/Myalgia

http://en.wikipedia.org/wiki/Arthralgia

http://en.wikipedia.org/wiki/Headache

http://en.wikipedia.org/wiki/Lymph_node



events for the world‘s first nuclear meltdown began at Chernobyl Unit 4 in Ukraine.

During a safety experiment designed to evaluate potential cooling-system backup

procedures, plant operators had disabled automatic warning systems, deactivated

automatic temperature and pressure-trip systems, and disregarded established plant

procedures. Sluggish responses and inattention by plant operators led to a situation in

which steam pressure in the reactor core rose uncontrollably, ultimately resulting in

an increase in reactor power to 100 times the design value, shattered fuel pellets,

ruptured fuel channels, a fire, and two explosions. Eight tons of fuel and other highly

radioactive materials were ejected from the reactor, and large quantities of hazardous

vapors were released into the atmosphere. The radioactive fallout was detected all

over the world, from Finland to South Africa, with the immediate aftermath in the

Soviet Union including 300 deaths, a $12.8 billion disruption to the economy, and a

host of problems for those living in the vicinity of the plant, including a doubling of

birth defects, a 5 to 10 times higher rate of thyroid cancer, and a 2 to 4 times increase

in leukemia rates among children in affected areas.

A similar but more immediately deadly disaster occurred in Bhopal, India, 2 years

earlier; however, this one involved chemical rather than nuclear contamination.

During the night of December 2, 1984, a toxic cloud of methyl isocyanate (MIC)

suddenly was released from the Union Carbide pesticide plant. No action was taken to

warn the surrounding population until more than an hour after the leak started. The

accident was immediately fatal to more than 2000 people living in the vicinity and has

since caused more than 300,000 injuries and as many as 6000 additional deaths.

While denying liability, Union Carbide agreed to a settlement of $470 million.

But the big question remains: What do all of these catastrophes have in common?

Most share complex interactions between engineering design issues, inadequate or

unclear operational processes, less-than-optimal staff competency, and insufficient

emergency response systems, among other factors. However, in each case, it is

noteworthy that the tragic chain of events involved one or more human errors that

occurred in the early morning hours when unintentional sleep episodes have been

shown to be more common. Although it cannot be proven with absolute certainty that

all these major industrial blunders resulted from sleepiness-related reductions in

alertness, decrements in cognitive processing, or poor reaction time, it is more than

coincidence that the disastrous mistakes were made at a time of day when human

functional capacity is known to be diminished [12].

3.5.1 COSTS ASSOCIATED WITH FATIGUE

Caldwell (2001) [12] continues in examining the relation between fatigue related

accident and their costs. In view of what is now known about fatigue, it is not

surprising that several studies have shown a significant increase in workplace

accidents during late night hours. Although studies aimed specifically at the aviation



sector are virtually nonexistent, conservative estimates from other occupational

settings suggest that the direct cost of fatigue in terms of injuries and loss of

production is at least $2 billion per year. However, the sum total of the overall effects

of sleep loss on work performance has been placed at $18 billion annually in the

United States alone. According to a recent survey, 40% of adults indicated they were

so sleepy during the day that it interfered with their daily activities, and 18% said they

suffered from this type of problem several days a week. Furthermore, more than half

the people surveyed also reported that chronic sleepiness adversely affected their

mood, energy levels, concentration ability, and overall health, as well as their ability

to pursue personal interests and maintain quality relationships with their family and

friends. Thus, fatigue touches every aspect of life in modern society, including air

EMS activities and other aviation sectors in which requirements for unpredictable and

extended work episodes often occur at times when alertness tends to be most

compromised. Several studies have tried to evaluate the costs to society of alertness

related accidents and loss of performance with one estimating costs exceeding 40

billion $ per year .

Figure 3. 13: Three Mile Island nuclear power plant

3.6 MEASURING FATIGUE

Sherry (2000) [87] states that measuring fatigue in the workplace is a complex

process. It is common to use both subjective and objective measures of fatigue and

alertness to evaluate the impact of a countermeasure, as multiple measures allow the

investigator to triangulate the truth and produce a more convincing conclusion. There

are four kinds of measures that are typically used in measuring fatigue; physiological,

behavioral, subjective self-report and performance measures.



The following parts were taken from Sherry‘s paper regarding fatigue in the railroad

industry where the author describes all four categories of fatigue measures [87].

3.6.1 PHYSIOLOGICAL MEASURES

Physiological measures that have been used to determine fatigue and sleep are the

Electroencephalograph (EEG) and the multiple sleep latency test (MSLT). The EEG

has been useful in determining the presence of ongoing brain activity. In general,

according to Monk (1991), ―the onset of sleep has been characterized by decreases in

fast, low voltage activity, decreases in the regularity and frequency of alpha activity

and increases in slow activities (delta and theta).‖ For the most part EEG measures

have been used as the reference point for calibrating other measures of sleep and

fatigue.

The Multiple Sleep Latency Test (MSLT) was developed by Carskadon and Dement

in 1977. The test measures the amount of time a test subject falls asleep in a

comfortable, sleep-friendly environment.

Figure 3. 14: The Multiple Sleep Latency Test

Another physiological measure that has been studied to detect fatigue has been eye-

movements. Several studies have looked at the effectiveness of eye movement,

(smooth pursuit, and saccadic) as indicators of sleepiness and fatigue. One study used

five subjects to complete the Multiple Sleep Latency Test and the Maintenance of



Wakefulness Test (MWT). Results showed that saccadic movement was significantly

correlated with increased sleepiness (Porcu, et al., 1998).

Another test used to indicate the presence of fatigue was the psychomotor vigilance

test (PVT) developed by Dinges and Powell (1985). This test uses a small yellow light

displayed on a computer screen. As soon as the light is displayed the test subject is

required to hit a key and stop the reaction time counter.

An increasingly popular method of detecting the presence of fatigue is the use of a

measure called PERCLOS (Percentage of Eye Closure). This measure attempts to

detect the percentage of eye-lid closure as a measure of real time fatigue. This

procedure is achieved by mounting a video camera directed towards the subject‘s eye.

As time progresses the individual‘s eyelid closure, rate of blinking, and degree of

closure is photographed. Subsequently, a variety of indices are computed. Several

studies indicate that PERCLOS has significant advantages over EEG algorithms in

detecting fatigue and drowsiness.

Figure 3. 15: Fatigue measurement tools. PERCLOS (left) and EEG (right)

A measure that has been studied recently, but for which inconclusive results have

been found is the Fitness for Duty (FIT) test by PMI, Inc. This test attempts to detect

fatigue through the measurement of the saccadic movements of the eye. Several

indices have been calculated such as the extent of smooth pursuit and the saccadic

movements.

3.6.2 BEHAVIORAL MEASURES

Behavioral measures of sleep have become popular over the last few years. The most

common behavioral measures are the actigraphs, devices that can measure sleep based

on the frequency of body movement. The test subject wears a wristwatch-like

recording device that detects wrist movements. The number of body movements

recorded during a specified time period, or epoch, has been significantly correlated



with the presence of sleep. Scientific results have shown a direct connection between

actigraph tests and that of the EEG.

Figure 3. 16: A common behavioral measure, the Actigraph

A recent study of NASA astronauts found that on a minute-by-minute basis, there was

a good correlation between sleep stage and actigraphic movement counts, with a

higher level of counts per minute recorded in epochs with lighter sleep stages. Results

showed that actigraphs perform properly at space providing very important data

regarding astronaut‘s sleeping patterns.

3.6.3 SELF REPORT MEASURES

In the scientific community a variety of self-report measures have been developed to

study fatigue, sleepiness, and alertness. These measures are easy to administer and

readily accepted by study participants. The Stanford Sleepiness Scale (SSS) (Hodes,

et al., 1973) consists of seven statements ranging from ―wide awake‖ to ―cannot stay

awake‖. The scale has been validated against performance measures as a function of

sleep deprivation and is perhaps the most widely used measure of subjective

sleepiness.

Another very widely used measure is the Visual Analog Scale (VAS). This type of

measure was initially developed for use in educational research (Monk, 1991) while it

has also been used in symptom measurement. VAS consists of a single horizontal line

presented on a piece of paper. At either end of the line an anchoring descriptor is

displayed such as ―wide awake‖ or ―about to fall asleep‖. Study participants are

instructed to place an X on the horizontal line between the anchors to indicate the

extent to which they are best description of their current state. The user-friendly

interface has made VAS very popular to study participants and researchers.



Figure 3. 17: The Epworth Sleeping Scale for sleepiness measurement

Another type of subjective self-report technique is the mood descriptors. The typical

measure of this sort is one in which a series of adjectives that indicate a variety of

different mood states are listed. Subjects must then check which adjectives best

describe their current state.

Finally, a common self-report measure has been the use of diaries. Several studies

have been conducted in which the participants were asked to keep detailed logs of

their activities including work and sleep. These diaries were then used to develop

descriptions of the sleep wake cycle of the participants and to develop a baseline for

further studies. A study of locomotive engineers conducted by Pollard (1996)

distributed diaries to over 300 engineers and conductors. While the return rate was

low, 24%, results indicated that railroad engineers averaged about seven hours and

eight minutes of total sleep per night.

Figure 3.18 shows a typical example of a sleeping diary. It is important to comment

that till now, individuals working in fatigue inducing environment have failed to

understand the significance of fatigue. This issue is reflected on the fact that few

people actually respond or take into serious consideration the importance of such

measures as sleeping diaries. Whether there was greater number of participants, the

findings that would be provided by this procedure would be an effective instrument

for better fatigue knowledge and its counter measuring.



Figure 3. 18: A typical sleeping diary

3.6.4 PERFORMANCE MEASURES

A common practice to determine the effects of sleep and sleep deprivation has been

the use of performance measures. Various tasks ranging from simple to complex have

been examined including reaction time to visual and auditory stimuli and vigilance

tests. For example, in one study, twelve male undergraduate students were deprived of

sleep for one night and were given a series of cognitive tasks. To explore the effects

of short-term sleep deprivation on attention, the following tasks were also

administered: a working memory task, a trail-making task, a vowel/consonant

discrimination task, and a letter recognition task.

Performance measures that have been used extensively in various studies include the

Walter Reed Performance Assessment Battery (PAB) (Thorne, 1985). This battery

consists of various measures of neurobehavioral and cognitive performance including:

two letter search, six letter search, encoding/decoding, two column addition, serial

addition subtraction, logical reasoning, digit recall, pattern recognition, pattern

recognition II, visual scanning, mood activation scale, mood scale two, four choice

serial reaction time and time estimation. In addition, several tests have been added to

the PAB: a visuo-spatial rotation task, interval production. Stroop, repeated

acquisition, repeated acquisition, code substitution, delayed recall, running, memory,

matching to sample, and the ten-choice reaction time (RT). All of these tests have



been demonstrated to have some sensitivity to the effects of sleep deprivation, jet lag,

heat stress, physical fatigue, physical conditioning, atropine, hypoxia, and sickle cell

disorders.

A variation on the PAB, adapted for Windows based operating systems is the Denver

Fatigue Inventory, a computer assisted cognitive test battery. The battery consists of

the choice reaction time test, the serial addition subtraction test, the manikin test, the

circle target test, and the light response test. These instruments have been used in

several studies with locomotive engineers and railroad dispatchers (Sherry, 1998).

Results from these uses have demonstrated significant differences between day and

midnight shifts. This review has demonstrated that there are a variety of different

techniques available for detecting fatigue, sleepiness, drowsiness, and the effects of

those conditions which can be sorted into several categories, (Sherry 2000).

However, as Sherry [87] mentioned:

―There is no objective tool to measure fatigue. It has been proposed that „fatigability‟

is an objective inability to sustain power, which can be measured by

electrophysiological methods, but to date, attempts to objectively measure fatigue

have failed.‖

One study compared single-photon emission computed tomography (SPECT) scans

between patients with chronic fatigue syndrome (CFS) and healthy persons when

performing attention and working memory test. There were no group differences for

the performance task, despite the fact that CFS subjects perceived them to require

more mental effort to perform the task.

Currently fatigue has typically been identified using a number of subjective scales.

More than 30 scales are available for measuring fatigue. Seven of the most frequently

used scales are described below [86]

3.6.5 OTHER FATIGUE MEASURMENTS TOOLS

Note: This excerpt was taken directly from [86] in order to present a brief review on

additional fatigue measurement tools.

Fatigue severity scale (FSS) this nine-item fatigue severity scale is one of the best

known and most used fatigue scales. The FSS items are principally measures of the

impact of fatigue on specific types of functioning, relating to the behavioral

consequences of fatigue rather than symptoms. The FSS has high internal consistency

with Cronbach‘s alpha between 0.81 and 0.89. Also, it is sensitive to change with time

and treatment, and has good test-retest reliability. The FSS is able to distinguish

patients with different diagnoses, such as between systemic lupus erythematosus

(SLE) and multiple sclerosis (MS), and between chronic fatigue syndrome (CFS), MS



and primary depression. Hossain et al. successfully used the FSS to identify the shift

workers with high fatigue from those with low fatigue.

Fatigue questionnaire (FQ): The FQ originally consisted of 14 items with a

subsequent revision to an 11-item questionnaire. The scale is comprised of two

dimensions: Physical Fatigue and Mental Fatigue. This scale was developed for

hospital and community studies with CFS and has been used in multiple studies. The

FQ was found to be both reliable and valid. There was a high degree of internal

consistency with a range of Cronbach‘s alpha of 0.88 and 0.90. The validity of the

FQ, in assessing fatigue, suggests that it is a useful tool for assessing fatigue in a

variety of medical disorders. It has been used to assess fatigue in patients with cancer

and HIV, and in other medical patients, and Gulf War veterans.

Multidimensional fatigue inventory (MFI-20): The MFI-20 is a 20-item self-report

instrument. These items rate the severity of fatigue in the past week. The inventory

covers 5 dimensions: General Fatigue, Physical Fatigue, Mental Fatigue, Reduced

Motivation and Reduced Activity. It has good internal consistency and test-retest

reliability. The MFI-20 was tested for its psychometric properties in cancer patients

receiving radiotherapy, patients with CFS, psychology and medical students, army

recruits and junior physicians. It was used to discriminate patients who had

Parkinson‘s disease from those who did not. The convergent validity of the MFI-20

was investigated by correlating the MFI subscales with a Visual Analogue Scale

measuring fatigue with the correlation co-efficient between 0.22 and 0.79.

Fatigue assessment instrument (FAI): The FAI is a 29-item fatigue assessment

instrument that includes four subscales: Fatigue Severity, Situation Specificity,

Consequences of Fatigue and Responsiveness to Rest/Sleep. The internal

consistencies of the corresponding subscales are good to excellent with Cronbach‘s

alpha values between 0.70 and 0.92.

The Fatigue Severity subscale has 11 items, eight of which correspond highly with

the FSS mentioned above, indicating a good convergent validity. Although its test-

retest reliability is only moderate, the inventory, in general, has good psychometric

qualities. The FAI is used to differentiate normal fatigue from medical disorders

commonly recognized to have a large fatigue component and is able to distinguish

differences between patients with different diagnoses.

Fatigue impact scale (FIS): This 40-item questionnaire is used to assess the impact

of fatigue on different functioning areas, such as cognitive, physical and psychosocial

functions. It has good internal consistency and reproducibility. The FIS was validated

in a subject sample of patients with MS and hypertension. There were significant

differences in the scores of these two groups. A function analysis was carried out in a

group of CFS patients and a group of MS patients. The results showed that the FIS

correctly classified 80.0% of the CFS group and 78.1% of the MS group when these



groups were compared. This validation study indicates that the FIS has considerable

merit as a measure of patients‘ fatigue symptoms. The FIS is also a useful instrument

to assess the impact of fatigue on patients‘ every-day lives.

The brief fatigue inventory (BFI): This nine-item scale was developed for screening

and assessing clinical outcome in fatigued patients with cancer and it identified those

patients with severe fatigue. The BFI is a reliable instrument. It is used for a rapid

assessment of fatigue severity in both clinical screening and clinical trials. The BFI

significantly correlated with two previously validated measures, the Profile of Mood

States (POMS) Fatigue subscale and the Functional Assessment of Cancer Therapy

(FACT), indicating that the BFI has good concurrent validity.

Figure 3.19 presents a depiction of the Fatigue Severity scale.

Figure 3. 19: Fatigue Severity Scale, (FSS) [86]

3.7 FATIGUE COUNTERMEASURES

Fatigue is a very common issue both in workplaces and domestic environments. The

first step for eliminating fatigue is recognizing the factors that create it. In the past

few years, many scholars and professionals from various scientific fields have tried to

battle fatigue through research and fatigue management. The recent trend, especially

in aviation, is companies and independent professionals that deal exclusively with

fatigue countermeasures and fatigue assessment. Their main goal is to reduce fatigue



through proper education and arrangements while their primary area of focus is work

shifts and work environments in general. However, fatigue is a rather obvious issue so

there are a lot of non-complex measures to reduce it.

Queensland government suggests five basic steps in the fatigue risk management

process (much similar to IMO‘s Formal Safety Assessment):

1. Identifying hazards

2. Assessing risks that may result because of these hazards

3. Deciding on control measures to prevent or minimize the level of risks

4. Implementing control measures and

5. Monitoring and reviewing the effectiveness of control measures.

The USDA Forest Service [105] proposes a rather similar approach:

Improve your fitness and maintain regular physical activity

Ensure appropriate rest before assignment or work shift

Practice work cycling (hard/easy, long/short)

Adjust your work to conditions (heat and humidity)

A very popular tool for fatigue mitigation is FAST. The Fatigue Avoidance

Scheduling Tool (FAST) was developed by the United States Air Force in 2000–2001

to address the problem of aircrew fatigue in aircrew flight scheduling. FAST is a

Windows program that allows scientists, planners and schedulers to quantify the

effects of various work-rest schedules on human performance. It allows work and

sleep data to be entered in graphic, symbolic (grid) and text formats. The graphic

input-output display shows cognitive performance effectiveness (y axis) as a function

of time (x axis). An upper, green area on the graph ends at the time for normal sleep,

90% effectiveness. The goal of the planner or scheduler is to keep performance

effectiveness at or above 90% by manipulating the timing and lengths of work and

rest periods. A work schedule is entered as red bands on the time line and sleep

periods are entered as blue bands across the time line, below the red bands.

The calculated performance effectiveness represents composite human performance

on a number of cognitive tasks, scaled from zero to 100%. The oscillating line in the

graph represents expected group average performance on these tasks as determined by

time of day, biological rhythms, time spent awake, and amount of sleep, and various

confidence limits around the average may be displayed. The graphic display may be

cut and pasted into reports and briefing slides. Cognitive effectiveness estimates for

work periods of any length may also be cut and pasted in tabular format. While FAST

began as an exclusive tool to the Aviation industry, a lot of FAST variations were

soon established in other fields of transport and industry in general [GG].

http://www.deir.qld.gov.au/workplace/subjects/fatigue/managing/risk/one/index.htm

http://www.deir.qld.gov.au/workplace/subjects/fatigue/managing/risk/two/index.htm

http://www.deir.qld.gov.au/workplace/subjects/fatigue/managing/risk/three/index.htm

http://www.deir.qld.gov.au/workplace/subjects/fatigue/managing/risk/four/index.htm

http://www.deir.qld.gov.au/workplace/subjects/fatigue/managing/risk/five/index.htm

http://en.wikipedia.org/wiki/United_States_Air_Force



Figure 3. 20: The Fatigue Avoidance Scheduling Tool

The field of fatigue countermeasures is a very rich area for discussion. Strategies for

mitigating fatigue in transportation will be thoroughly discussed in the next chapter

while our overall perception of the situation and our possible recommendations will

be part of the last chapter of this thesis.

In figure 3.21 we provide a summarized representation of the most common fatigue

countermeasures in accordance to their effectiveness and their apparent feasibility. It

is that certain practices, such as the night/mourning shift rotation, cannot be avoided.

On the other hand, listening to music/radio may not be as effective as sufficient sleep

but can mitigate fatigue and increase alertness.



Figure 3. 21: Effectiveness, feasibility and duration of effect for common Fatigue

Countermeasures



CHAPTER 4 FATIGUE IN TRANSPORTATION



4 FATIGUE IN TRANSPORTATION

4.1 GENERAL OVERVIEW

It is undeniable that fatigue is widely regarded as one of the contributing factors to

transportation accidents. The exact impact of fatigue on these incidents remains, even

today, a rather difficult and complex task. However, no one can deny its significance

although it is usually underestimated. The presence of fatigue in transport has led to

thorough research, especially during the past two decades. The main areas of focus

are two: a) the issue of fatigue in terms of physiological measures and b) the actual

measurement and quantification of fatigue in terms of levels. Unfortunately, the

physiological fatigue mechanism has proven to be extremely complex to scientists

since it is the result of a plethora of different interacting parties.

Smith, Allen and Wadforth [90] indicate that there is a rather long history of

investigating impact of fatigue in other transport sectors. Fatigue investigation has

been developed from three different fields, as mentioned below:

Empirical reports about the impact of fatigue

Laboratory research about fatigue severity and outcomes

Extensive research and experiments in military transportation

These types of military transport research have been the first step into a series of

related studies, most of whom are solely addressed to driving. Focusing on driver‘s

fatigue seems only logical since driving is regarded as a crucial part of public safety

rather than just an issue of a general occupational framework.

Smith et al [90] refers to several International meetings (see Hartley, 1997, Akerstedt

and Haraldasson, 2001) that have provided interesting overviews of the area and

developed a framework for evidence-based countermeasures. The main problem that

has been addressed is that fatigue has severe impact on all areas of transport, an

impact which has been overlooked in the past, (Akerstedt and Haraldasson, 2001).

While a lot of regulations and guidelines have been established and made obligatory,

none of them achieves to include all eight fatigue related criteria. Those criteria are

mostly related to working hours and work environment.

Smith [95] pinpoints a proposition made by many parties for fatigue to be dealt with a

hybrid approach incorporating both a prescriptive ―hours of service‖ system and a

non-prescriptive, outcomes-based approach.

A general overview of research made on transport fatigue can be found on papers

presented at the 2005 International Conference on Fatigue Management



Transportation Operations .The papers show the range of issues being studied, issues

such as fatigue on fundamental skills required in transport operations or

epidemiological studies of fatigue; evaluation of countermeasures; and assessment of

fatigue management programs.

4.1.1 FATIGUE IN AVIATION

Due to its complex and demanding nature, aviation has been the spark for many

studies regarding fatigue. Statistics indicate that human fatigue has been a

contributing factor in a great number of severe aircraft accidents. Generally, it has

been estimated that fatigue contributes to 20-30% of transport accidents. In

commercial aviation operations, about 70% of fatal accidents are related to human

error, therefore the risk of crew fatigue contributes to about 15-20% of the overall

accident rate.

Some aircraft accidents that were found to involve fatigue:

1993 Kalitta International, DC-8-61F at Guantanamo Bay, Cuba

1997 Korean Air, 747-300 at Guam

1999 American Airlines, MD-82 at Little Rock, AR

2004 MK Airlines, 747-200F at Halifax, Nova Scotia

2004 Corporate Airlines, BAE Jetstream31 at Kirksville, USA

2004 Med Air, Learjet35A at San Bernadino, CA

2005 Loganair, B-N Islander at Machrihanish, UK

2006, 27th Aug, Comair, CRJ100 at Lexington, KY

2007, 25th June, Cathay Pacific 747F at Stockholm, Sweden

2007, 28th Oct, JetX, 737-800TF-JXF Keflavik airport, Iceland

According to Smith et al [90], great concern with fatigue in aviation developed during

the Second World War. The NASA-Ames research group has been a pioneer in

investigating flight crew fatigue in commercial pilots (Gander, Graeber, Connell,

Gregory, Miller and Rosekind, 1998). The area of focus of these projects is mostly

circadian disruption, sleep quantity and fatigue before and at the end of flights. As

mentioned in Chapter 3, Circadian rhythms are highly affected by jet lag so it is only

logical that circadian disruption is a very common issue among flight crews.

In order to present answers to aviation related fatigue, fatigue risk management

systems have been developed. Apart from FAST which we discussed earlier Canadian

aviation presented the ‗Fatigue Risk Management Toolbox‘, a very practical tool,

which typically consists of:

Policy templates and guidelines to assist in the development of global and

detailed corporate policies on the management of fatigue.



Competency-based training and assessment for employees, management and

new staff.

Fatigue audit tools to assess work schedules, verify actual fatigue levels and

monitor the fatigue risk management process

Figure 4. 1: The air crash of American Airlines, MD-82 at Little Rock, AR

4.1.2 RAILWAY OPERATIONS FATIGUE

The investigation of fatigue in railway operations has been an active area of research

for over five decades. While the general approach of researchers bears a lot of

similarities to that regarding drivers-fatigue, the main focus is the interaction of

fatigue with specific incidents such as signal misses (Buck and Lamonde, 1993). A

series of studies from all around the globe (e.g. Poland - Malgarzeta, 1982; China -

Zhou, 1991) have made clear that fatigue is present in the railroad industry and its

effects are to not to be neglected [90].

The recent establishment of the Federal Railroad Administration‘s Fatigue Research

Program (Sussman and Coplen (2000) and Pilcher and Coplen (2000) has indicated

that risk factors in railways are mainly overall sleep quality and lack of rest breaks.

The program‘s overall approach suggests better data collecting techniques, more

sufficient measurement and analysis tools and finally proper and applicable ways of

mitigating fatigue.

A rather practical and effective approach for railway industry fatigue was the

development of the HSE Fatigue index (Spencer, Robertson and Folkard, 2006)



[35].The fatigue risk factors were recorded via diary entries. Moreover, the research

team tried to connect these entries to specific incidents that would cause an accident.

At the end of the study Spencer, Robertson and Folkard recommended:

1. Shorter shift duration

2. A maximum of four hours of non-stop

3. Maximum work time during a week

4. Reduction of continuous night shifts

5. Obligatory rest periods

6. Rest breaks between shift changes.

Smith et al [90] agree that the HSE Fatigue Index is probably the most effective

option for the fatigue management of rail workers. Consequently, this will have a

great impact on forthcoming legislations as a recommendation made by HSE will be

taken into consideration so that fatigue levels are reduced in the ranks of rail workers

and shift workers in the transport sector in general.

4.1.3 FATIGUE IN DRIVING

Recent studies by the National Transportation Safety Board (NTSB) have indicated

that fatigue resulting to sleepiness is a major factor in accidents involving heavy and

commercial vehicles.

According to smartmotorist.com [EE] the NTSB states that 52 per cent of 107 single-

vehicle accidents involving heavy trucks had fatigue playing an important part in their

happening. Reports made clear that in nearly 18 per cent of the incidents, the driver

actually fell asleep. According to data from US Department of Transportation, the

percentage of fatigue-related fatal accidents is estimated to be around 30%. More over

statistical data from USDT show that driver fatigue is responsible, among many

factors, for approximately 20% of commercial road transport crashes and over 50% of

long haul drivers have fallen asleep at the wheel. Recently The National Highway

Traffic Safety Administration (NHTSA) estimate that there are 56,000 sleep related

road crashes annually in the USA, resulting in 40,000 injuries and 1,550 fatalities.

Several studies in the US have gathered interesting statistics regarding fatigue on

the road [EE]:

50% of fatal accidents on those roads were fatigue related

that 30% - 40% of accidents involving heavy trucks are caused by driver

sleepiness

In 2002 alone the Total Cost of Fatigue-Related Crashes exceeded $2.3 billion

The presence of fatigue can cause a serious impairment of an individual‘s driving

skills. High fatigue levels can make reacting much slower, distract concentration on



driving and make the surrounding interpretation a difficult task to perform. Fatigue

related crashes tend to be more severe because of the limited time available for

reaction in order to avoid the accident. The combination of fatigue and alcohol

consumption multiplies the probability for a lethal road accident.

Over the period of 2002 to 2006, driver fatigue was identified as being responsible for

256 deaths or twelve percent of fatal crashes, and more than 4,350 injuries in the US.

These numbers have convinced scholars and government to address the issue more

thoroughly. Recent international research has suggested, due to fatigue under-

reporting, the percentage of drivers-fatigue related incidents is significantly greater

than it is reported to be. Some rough estimation suggests that the actual percentage is

more than double.

The most common signs that can alert a driver are:

- Frequent blinking or yawning

- Difficulty to keep your head up

- Eyes closing

- Inability to focus concentration solely on the road

- Braking too late

- Drifting over the median line unto the other side of the road (which is actually the

most alerting sign of all the above).

Figure 4. 2: A driver about to experience a microsleep

Smith et al. [90] suggests that the absence of supervision during long distance

transport has allowed fatigue to run more freely on the highways. Many studies (e.g.



Landstrom, Akerstedt, Bystrom, Nordstrom and Wiborn, 2004) insist that rest breaks

and napping are highly recommended to battle fatigue on the road.

Several countries have also convened expert panels to review regulatory options for

reducing heavy vehicle driver fatigue (National Road Transport Commission, 2001;

Transport Development Centre, Transport Canada, 1998; University of Michigan

Transportation Research Centre, 1998) [90].

4.2 LEGISLATION REGARDING FATIGUE IN TRANSPORT

There are several regulations and guidelines regarding fatigue avoidance and

mitigation. Jones, Dorrian et al. (2005) [44] have provided some collective figures

regarding the issue, in different areas of transportation.

In figure 4.3 we see the first, most significant and most recently revised regulations

regarding fatigue and working hours.

Figure 4. 3: The first, most significant and most recently revised regulations [44]



Most industries have tried to a establish work/rest patterns in order to eliminate the

presence of fatigue in the workplace. Smith et al [90] summarize the most common

issues that these regulations address:

• Sleep: Minimum hours of sleep per day with emphasis on sleep quality, continuity

and potential deprivation. The issue of circadian rhythms and disruption is also

mentioned by almost all regulations as an integral factor for maintaining alertness

levels.

• The cumulative nature of fatigue and sleep loss: Repetitive sleep deprivation

should be avoided. The cumulative nature of fatigue increases the danger of severe

alertness drops.

• Recovery sleep: Recovery sleep is essential after long periods of sleep deprivation

in order to recharge the sleep batteries.

• Night work: The human body is not intended to be awake at night. This fact

increases the danger of reduced alertness on night shifts even if an individual work at

night on a regular basis.

• Rest breaks: There should be an equivalent between actual working time and rest

time. Most regulations suggest about 10 % of the working time.

• Duration of working time: There must be a maximum working time per day or

week. Otherwise the danger of fatigue is inevitable.

Many countries advocate the use of fatigue management programs that consist the

following:

• Organizational commitment to the requirements of a ‗Fatigue Management

Program‘

• Establishment of a ‗Fatigue Management Policy and Process‘

• Involvement of all stakeholders throughout the process

• Competency based educational modules

• Effective change to the scheduling, dispatching and compensation processes.

• Objective and subjective measures of fatigue management effectiveness.

• Continual monitoring and improvement [90].

In figure 4.4 provided by Jones et al. (2005) [44], there is a collective representation

of how the legislation parties have addressed the eight fatigue related criteria which

that were discussed in Chapter 3.



Figure 4. 4: Legislation regarding Fatigue in Transportation [44]

4.3. FATIGUE IN MARITIME TRANSPORT

In the previous chapters the connection between fatigue and various fields of transport

was thoroughly examined. Recent studies have shown that fatigue is present in the

maritime industry as well.

Hetherington, Flin and Mearns (2005) [31] state that contemporary research has

shown that there are potentially disastrous outcomes from fatigue in terms of health

and performance. The maritime industry has not yet forgotten the grounding of Exxon

Valdez in 1989, one of the most terrible environmental disasters in the history of

transportation. The incident report shows that in the 24 hours prior to the grounding,

three watchkeepers had only had 5 or 6 hours of sleep (National Transportation Safety

Board [NTSB] 1990), suggesting that fatigue may have been a contributing factor to

this unfortunate catastrophe. But even prior to the Exxon Valdez grounding, fatigue

was also present at sea.

In the wake of the twenty first century, the seafarers work has become more

demanding than ever due to a combination of minimal manning, sequences of rapid



turnarounds and short sea passages, adverse weather and traffic conditions, long

working hours with insufficient opportunities for recuperative rest [90].

Despite the introduction of work rest mandates by the IMO (2001) and the STCW

(1995) many seafarers have to work for more than 12 hours a day with a maximum of

5-6 hours break. A very representative example of an officer‘s lack for rest time is the

fact that during the discharging of a VLCC the chief officer must be present at all

times. A VLCC with a 300,000 tonnage takes approximately 44-46 hours to

discharge, so this means that the chief officer is required to be awake and present for

almost two days time [31].

Furthermore, in a report by the National Transportation Safety Board (1999) seafarers

were identified out of the occupational groups included to be second in highest

number of continuous working time, close behind rail workers.

Moreover, the seafarers themselves recognized that the impact of fatigue has been

significantly greater in the past 10 years than it used to be. Out of 1000 participating

officers, 84% felt that stress was also more prevalent,(Cole-Davies 2001). Further

studies conducted my NUMAST (National Union of Marine Aviation and Shipping

Transport Officers), MCA (Maritime Coastguard Agency), AMCA and MAIB

respectively have presented the following foundings:

70% of seafarers report poor to very poor sleep

fatigue related accidents were most prevalent at the beginning of the tour (first

week), in the first four hours of a shift, between the hours of 09:00 and 16:00,

and in calm conditions [93]

66% felt that extra manning was necessary

50% of which indicated that they worked more than 85 hours in a week.

Dr. Rothblum [80] of the US Coast Guard indicates that the NTSB has identified

fatigue to be an important cross-modal issue, being just as pertinent and in need of

improvement in the maritime industry as it is in the aviation, rail, and automotive

industries. In addition, several studies have cited fatigue as the primary concern of

mariners. A US Coast Guard survey has also shown that fatigue contributed to 16% of

the vessel casualties and 33% of the injuries.

4.3.1 MARITIME FATIGUE RISK FACTORS

In the 2001 submission to the IMO (MSC 74/15, 2001), Guidelines on Fatigue

provided practical information to assist the marine industry in understanding and

mitigating fatigue.

In these guidelines, the potential causes of fatigue were categorized as follows:

Crew-specific



Physical Environment

Ship-specific

Management

The clear categorization of fatigue risk factors has proven very useful in investigating

fatigue in various types of vessels and crews. McCafferty and Baker (2002) [58]

indicate that through the work of many different segments of the maritime industry,

the variety of factors which impact crewmember fatigue and its effects on human

performance, are becoming better recognized [58].

4.3.1.1. CREW SPECIFIC FATIGUE FACTORS

The crew specific fatigue factors have a major impact on human performance and

reliability. Human error is one of the most recognized factors that lead to maritime

accidents.

According to the IMO (2001) [36] crew specific factors include:

Sleep and Rest

Quality, Quantity and Duration of Sleep

Sleep Disorders/Disturbances

Rest Breaks

Biological Clock/Circadian Rhythms

Psychological and Emotional Factors, including stress

Fear

Monotony and Boredom

Health

Diet

Illness

Stress

Skill, knowledge and training as it relates to the job

Personal problems

Interpersonal relationships

McCafferty and Baker (2002) [58] pinpoint three conditions that play an integral part

in the increase of fatigue levels, which are mentioned below:

A) Conditions of Sleep.

• Ability to mass sleep for a period of at least eight hours

• Sleep comfort

• Amount of daily sleep.



B) Biological Clocks/Circadian Rhythms.

Circadian disruptions are very common at sea since crew members normally change

shifts (night/day) throughout a trip thus disturbing the normal function of their

biological clock.

C) Workload and Work Duration.

• Cognitive and physical energy expended during task performance and over time

• Provision of brief rest periods during the work day

• Length of the workday

Other crew specific factors that affect alertness and performance according to the

IMO [34] are:

Age

Health

Diet

Substance use

Marital Status

Several studies have indicated that sleeping quality is by far the most important crew

specific factor for the prevalence of fatigue. The following figure shows the result of a

survey in which the participants were Australian pilot in the Great reef area. It is more

than obvious that sleeping at sea is significantly harder than in land.

Figure 4. 5: Sleep difficulties at sea and disturbances, ashore and at home.



4.3.1.2 SHIP SPECIFIC FATIGUE FACTORS

According to IMO (2001) [34] ship specific factors include:

Ship design

Level of automation

Level of redundancy

Equipment reliability

Inspection and maintenance

Age of vessel

Physical comfort in work spaces

Location of quarters

Ship motion

Physical comfort of accommodation spaces.

Another group of factors which have an impact on the mariner‘s performance and

fatigue is the overall design of the vessel itself. McCafferty and Baker (2002) [58]

divide these design factors in two subcategories.

1) Human-machine interaction aspects

2) Accommodation and overall work environment

The quality of a vessel‘s overall design is perhaps the more profound way for a naval

architect to improve the on-board work environment, thus indirectly mitigating

fatigue. ABS studies has shown by providing working and living spaces where

human factors/ergonomics principles are applied, fatigue can be controlled or even

eliminated while providing a healthy, safe and performance enhancing environment

for the seafarers. The IMO guidelines [34] propose that ship specific factors can be

assessed by emphasizing in habitability and in other general ergonomic principles.

The primary objective of ergonomics is applying knowledge and technology in

creating safe systems and healthy environments, reducing people‘s limitations and

enhancing their mental and physical capabilities. Moreover, this objective can be

attained by establishing regulations and guidelines that promote high compatibility

between ship design and mariner‘s performance. The ergonomic design of ships,

marine structures, and equipment is addressed in numerous handbooks, textbooks,

guides, and standards. In the past two decades some of the documents that addressed

the maritime industry in terms of design and ergonomics are:

• IMO's Guidelines on Ergonomic Criteria for Bridge Equipment and Layout (MSC

73/Circ 982)

• IMO's Navigational Bridge Visibility and Functions (IMO Res.A.708-(17)



• Ship‘s Bridge Layout and Associated Equipment – Requirements and Guidelines

(BS EN ISO 8468)

• IMO's Role of the Human Element in Maritime Casualties - Engine Room Design

and Arrangements. (IMO DE 38/20/1)

• IMO's Guidelines for Engine Room Layout, Design and Arrangement (MSC 68/Circ

834)

• IMO's Role of the Human Element in Maritime Casualties - Guidelines for the On

Board Application of Computers (IMO DE 38/20/2)

• ABS Guide for Crew Habitability on Ships (2001)

• ASTM's Standard Practice for Human Engineering Program Requirements for Ships

and Maritime Systems, Equipment, and Facilities (F-1337) (1991)

• ASTM's Standard Practice for Human Engineering Requirements for Ships and

Maritime Systems, Equipment, and Facilities. (F-1166)(2000) [58].

The American Bureau of Shipping has been very active in the field of ship design

ergonomics by providing papers and guidelines regarding the issue. One of the most

interesting and innovative publications was ABS‘s Guidance Notes on the Application

of Ergonomics to Marine Systems (1998). By pinpointing the interplay between ship

design, human error and fatigue ABS managed to show the shipping industry the

importance of improving working and living conditions at sea.

Calhoun [14] in collaboration with the University of Michigan‘s School of Naval

Architecture and Marine Engineering, emphasized in six specific factors that have a

direct impact on on-board performance:

Light

Noise

Vibrations

Ventilation

Temperature

Ship motions

Figure 4.6 provides a brief representation of factors that can lead to efficient vigilance

while highlighting those that Naval Architecture has a great chance to improve. As it

can be easily seen the role of the Naval Architect can be beneficial to several parts of



the overall procedure. Designing factors influence three different parts of the entire

environment including sleeping, working and watch station.

Figure 4. 6: Vigilance though improved ship design [14]

By focusing on these factors the Naval Architect can improve the ship design

providing an upgraded working and living environment. While most of these factors

can be assessed by various areas of engineering, vibrations and ship motions are

exclusive to naval architects as they rest upon ship design elements (hydrostatics,

hydrodynamics and hull structure).



4.3.1.2.1 LIGHT

Light is an integral part of an ergonomic environment. While the importance of

lighting is rather obvious as far as visibility is concerned, its primary purpose is to

maintain the regular pace of the circadian rhythms. Solar light is essential in

calibrating our biological clock, making the exposure to natural daylight a great need.

Since most mariners have to work night shifts, this need is not satisfied in a regular

basis resulting in casual circadian rhythms disruption. The most important aspect of

this disruption is the increase of drowsiness time of an individual and consequently

the probability of an error. Even in the morning hours, several crew members spend

the majority of their day within the confines of the vessel and below decks, where

electric light is all that is available. This can create an irregular and shifting sleep

cycle that leads to fatigue. Surveys conducted on mariners have shown that the

lighting systems installed aboard ships fail to be stimulating and maintain high

alertness levels on crew members.

Figure 4. 7: The lack of sunlight in a vessel's Engine Room

Even though there are several maritime lighting guidelines, most of them neglect the

importance of proper lighting in the biological clock circle. Moreover, they focus on



general aspects of functionality but fail to prevent the drop of alertness and the

prevalence of fatigue.

Artificial light may never be a full substitute of natural sunlight but high intensity

lighting has proven to be an effective means to stimulate fatigued operators and help

improve human performance by increasing alertness. Scientists insist that by

providing 1000 lux or more, the effects of fatigue may decrease.

The guidelines currently used by class societies are specifically task oriented, as

shown in figure 4.8. The illumination levels required to shift circadian rhythms and to

have a stimulating effect on fatigued operators (~1000 lux) are greater than most of

the values recommended by ABS [14].

Table 3 is an excerpt from the American Bureau of Shipping (ABS) "Guidance Notes

on the Application of Ergonomics to Marine Systems."

Figure 4. 8: ABS Lighting Guidelines [14]



4.3.1.2.2 NOISE

The common definition of noise is the presence of undesirable sound. On-board, noise

can be the result of various procedures and activities. Its presence can be found in

almost every compartment of the vessel making it difficult to entirely eliminate it.

Some of the most regular noise-related areas of the ship are engines, pumps and other

mechanical equipment.

Noise can affect both physical and mental human condition by inducing fatigue on the

workplace and decreasing human effectiveness. Apart from the working environment,

noise has been recorded as one of the most integral factors that affect sleep quality.

According to Calhoun [14] guidelines and regulations about noise tend to neglect its

impact on fatigue inducing since they focus on the prevention of potential hearing

loss. Temporary loss of hearing can be the result of even short-term exposure to noise

and can also lead to permanent hearing loss. Moreover, guidelines about noise fail to

address important physiological outcomes that may appear years after working in a

noisy workplace.

Medical experts indicate that the human body perceives all noise as a potential threat

or warning, so it responses in a defensive way. Even if noises coming from a piece of

equipment are not always a sign of warning, the human body responds as if they

actually were. Surveys on mariners have shown that people who work in a noise-full

environment are expected to have mood swings, short temper and inability to deal

with minor frustrations.

Calhoun [14] describes four effective Noise Control Methods as shown below:

Isolation via absorbing material

Noise Barriers that block transmission

Damping eliminating vibration impact

Absorption of noise and converting it to heat

ABS through its guidelines regarding noise levels makes recommendations about

maximum and preferred levels of dBs (decibels). Interviews and inspections have

shown that these levels are often surpassed (figure 4.9).

4.3.1.2.3 SHIP MOTION INDUCED FATIGUE

Motion-induced fatigue as part of naval biodynamic problems is a significant

contributor to the performance of ship crews (Coldwell 1989). Coldwell also

suggested that this was an important matter for the naval community and which

implied a higher incidence of mistakes, some of which may not be easily noticed by a

supervisor [23].



Figure 4. 9:ABS Guidelines regarding Noise levels aboard ships [14]

Working in a constantly moving environment can often induce muscle fatigue since

the vessel‘s motions would probably raise the average energy expenditure of the crew

members. This can be easily explained by the need of a crew member to maintain his

position/stance while performing a specific task at the same time.

Dobie [23] suggests that whole-body vibration may have impact on general comfort

and performance, while in worst cases in health and safety. Moreover, Dobie

continues by indicating that whole body vibration can be transmitted to the body in

three types of ways:

The whole body surface

The buttocks

By several individual parts of the body



Figure 4. 10: Transmission of vibration throughout the human body

Apart from the human body, vibrations can have a critical effect on the visual target

of a crew member (e.g. a control panel), thus resulting in decision making and

interpretation errors.

Most common sources of these vibrations, apart from the vessel motions, are these

caused by large pieces of equipment such as the propeller and the main engine. Dobie

pinpoints that higher frequency vibrations can also originate from hull responses

caused by severe slamming in heavy seas.

Calhoun [14] indicates that maritime vibration guidelines, in the same manner to

those about noise that were discussed earlier, mainly focus on preventing bodily

injury but neglect the impact of vibrations on fatigue and sleep quality. Some effects

of whole body vibration are listed below:

Physiological:

Cardiac rhythm increases

Respiration rhythm increases

Blood circulation increases

Vasoconstriction

Endocrine secretions

Central nervous system affected



Comfort and Performance:

Pain

Nausea

Vision problems

Posture

Movement and coordination decline

Force

Perceptions altered

The above effects are sometimes imperceptible to the operator. They also occur at

lower vibration levels than those currently established by vibration guidelines [14].

A lot of research has been conducted regarding the mitigation of vibration on-board.

Most of them propose practical countermeasures similar to those proposed for noise.

The use of isolators, absorption techniques and materials that can minimize the

frequency and intensity of vibrations is highly recommended.

Vibration noise transmitted into the structure of a vessel is best prevented through the

isolation of the machinery from the hull. Rubber padding and spring or rubber mounts

are some of the measures used for preventing this. The primary goal is to isolate

vibrations and noise to the engine room and stopping them from reaching the hull

structure.

In their quest of diminishing the unwanted effects of noise and vibrations, naval

architects should be aware of some drawbacks that come along. The added weight and

cost of suppression materials are a significant issue that impedes the designer from

integrating these measures in the vessels design. While IMO has been more than

active in the past 10 years in the area of Human Factors, the classification societies

have not addressed the issue adequately. It is rather clear that there is a great need for

more research on vibration induced fatigue and the human element in general.

4.3.1.3 PHYSICAL ENVIRONMENT FATIGUE FACTORS

According to IMO (2001) [34], physical environment factors can be summarized in

the following text:

“Exposure to excess levels of environmental factors, e.g. temperature, humidity,

excessive noise levels, can cause or affect fatigue. Long-term exposure may even

cause harm to a person‟s health. Furthermore, considering that environmental factors

may produce physical discomfort, they can also cause or contribute to the disruption

of sleep.”



Ship motion, noise and vibrations which were discussed in the previous paragraph are

the main contributing environmental factors. Within the ABS Guide to Crew

Habitability on Ships, habitability criteria and measurement methodologies are

presented for optimizing conditions of the ambient working and living environment.

Relevant information for vibration, noise, indoor climate, and lighting are given in

order to control fatigue related factors of the physical environment [58].

A typical example of the ABS Guide to Crew Habitability on Ships [61] regarding

vibrations can be seen in figure 4.4.

Figure 4. 11: Vibrations effects to human body and movement [61]

4.3.1.4 MANAGEMENT FATIGUE FACTORS

The last category of factors resulting in fatigue is that related to management issues.

According to the IMO (2001) [34] these factors include:

1. Organizational Factors

• Staffing policies and Retention

• Role of riders and shore personnel

• Paperwork requirements

• Economics

• Schedules-shift, Overtime, Breaks

• Company culture and Management style

• Rules and Regulations

• Resources

• Upkeep of vessel

• Training and Selection of crew

2. Voyage and Scheduling Factors



• Frequency of port calls

• Time between ports

• Routing

• Weather and Sea condition on route

• Traffic density on route

• Nature of duties/workload while in port

These factors are recognized to influence on-board fatigue and are mostly related to

management policy and practice. Examples of these factors include top-level

organizational goals, shipping companies regarding decisions relating to vessel

operations, schedules, and voyages and compliance to regulations.

In paragraph 4.5 where legislation regarding fatigue in the maritime industry is

discussed, these factors will be presented more thoroughly in accordance to the

STCW, IMO and various relevant regulations.

4.3.2 THE IMPACT OF STRESS ON MARINER’S FATIGUE

Stress is a contributory factor to the productivity and health costs of an organization

as well as to personnel health and welfare (Cooper, Dewe, & O'Driscoll). A survey by

Parker et al. (1997) for AMSA (Australian Maritime Safety Authority) presented

some very interesting results regarding seafarer‘s stress. By the use of a self report

questionnaire, respondents were asked to rate how frequently they felt stressed and at

what level. Moreover, they were asked how frequently and to what extent did they

engage in health related behaviors such as exercising and drinking.

The number of participants was 1,806 seafarers including crew, masters, mates, pilots,

and engineers. In addition, the questionnaire was given to a normative group of

people in order to compare the results to those of the mariners. Seafarers reported

significantly higher levels of stress from sources of work pressure than did the

normative group, especially when asked about relationships with others and the

home/work interface.

The general results indicated that:

80% seafarers reported occasional to frequent stress at sea over

65% of engineers, 60% of crew, and over 60% of masters reported moderate

to high stress levels.

In the same study the interaction between stress and decreased levels of alertness was

highlighted indicating that chronic psychological stress – the type of stress induced by

interpersonal relationships, task design, and management style – creates a constant

drain on crewmember energy levels [31].

http://www.amsa.gov.au/



The Crew Endurance Management (CEM) [103] guide by the US Coast Guard list the

most common cause for stress as:

Interpersonal relationships:

Lack of support from coworkers and supervisors

Task design:

Heavy workload

Infrequent rest breaks

Long work hours

Shiftwork

Hectic routine tasks

Little sense of control

Management style:

Authoritative management style

Lack of participation by workers in decision making

Poor communication between management and employees

Ambiguity or conflicting requirements

Lack of family-friendly policies

4.4 RESEARCH REGARDING FATIGUE IN MARITIME TRANSPORATION

The past two decades a lot of parties have been active in investigating, measuring and

battling fatigue in the maritime domain. These parties include government

organizations, classification societies, private researchers and scholars. Some of these

studies and surveys are listed below:

4.4.1 ITF SEAFARER FATIGUE: Wake up to the dangers (1997)

The participants of the ITF report [V] were 2,500 seafarers of 60 nationalities, serving

under 63 different flags. The results of the survey indicate that fatigue at sea is most

of the times the long hours that crew members have to work during their time on-

board. Some of the results showed:

65 % of the participants stated that their average working hours were more

than 60 hours per week

25% of them reported working more than 80 hours a week (42% of masters)

36% of the sample was unable to regularly obtain 10 hours rest per day

18% regularly unable to obtain a minimum of 6 hours uninterrupted rest.



55% considered that their working hours presented a danger to their personal

health and safety. Indeed, nearly half the sample felt that their working hours

presented a danger to safe operations on their vessel.

60% reported that their hours had increased in the past 5 to 10 years.

Fatigue was more evident during the first hours of their duty.

Long tours of duty were also common (30% reporting usual tour lengths of 26

weeks or above).

4.4.2 THE NEW ZEALAND MARITIME SAFETY

This report summarizes some of the most important risk factors that contribute to

fatigue by accumulating data from seafarers in New Zealand. Some of the most

important findings of this report include:

60% of seafarers admitted falling asleep on duty at least once

Most of the participants indicate that the reduction of manning levels and other

economic/management factors resulted in fatigued operators

The survey of masters and mates on the Cook Straight ferries found that the

key cause of fatigue was shorter more disrupted sleep

Ship motions, noise and vibration were also recognized as fatigue related

factors

Workers had to work in most of the scheduled rest

The average age of the crew was significantly high (above 50 years).

4.4.3 THE CARDIFF PROGRAMM

In 2006 Smith, Allen and Wadsworth, introduced the The Cardiff Seafarers‟ Fatigue

research programme [90]. Their main areas of focus were:

Prediction of the worst possible fatigue scenario for health and safety

Practical recommendations for mitigating fatigue in accordance to the vessel

type and trade

Advice packages in order to battle fatigue by recognizing hidden hazards in

working environments, work shifts and others aspects of life at sea.

The realization of this project was successful by using the following elements:

• Reviews of the literature

• A questionnaire survey of working and rest hours, physical and mental health

• Physiological assays assessing fatigue

• Instrument recordings of sleep quality, ship motion and noise

• Self-report diaries recording sleep quality and work patterns

• Objective assessments and subjective ratings of mental functioning



• Analysis of accident and injury data [90]

4.4.4 MAIB BRIDGE WATCHKEEPING SAFETY STUDY (2004)

The Marine Accident Investigation Branch, in their bridge watchkeeping safety study

[54] also recognized fatigue in their investigation of collisions and groundings of

several vessels due to errors occurring on the bridge. The study shows that twelve

vessels grounded with sole watchkeepers. All were dry cargo/container vessels, 84%

were less than 3,000 GT (Gross Tonnage), 92% carried only two deck officers, and all

had three or fewer deck ratings. All of these groundings occurred in clear or good

visibility, and 75% occurred during darkness. Only one grounding occurred in a port

or harbor area, the remainder occurred during coastal passage.

This data highlights the link between small dry cargo ships operating in the short sea

and fatigue induced by insufficient manning. The same study indicates that alertness

and performance tend to be at their lowest during the early hours of the morning, a

fact well discussed earlier in this thesis.

MAIB have concluded that that the records of hours of rest on board many vessels,

which almost invariably show compliance with the regulations, are not completely

accurate. The number of recent groundings in which fatigue has been a contributory

factor, indicates that the hours of rest regulations are not having significant effect with

regards to the bridge watchkeeping arrangements on many vessels. There is also

pressure on masters and chief officers to do some of their ancillary work while on

watch, with the inevitable consequence of degraded attention to their watchkeeping

duties [54].

4.4.5 THE TNO REPORT (2005)

In their 2005 report on fatigue in the shipping industry, Houtman et al (2005) [33]

made a thorough presentation of the issue by recognizing all the parties involved (e.g.

shipping companies). TNO research on other sectors of transport was critical in

recognizing dangers that were till then overlooked.

Focus was given on twelve areas for fatigue management:

• Lengthening of the resting period

• Optimizing the organization of work

• Reducing administrative tasks

• Less visitors / inspectors in the harbor / better co-ordination of inspections

• Reducing overtime

• Proper Human Resource Management

• Education and training

• Development of a management tool for fatigue



• Proper implementation of the ISM-code

• Healthy design of the ship

• Health promotion at work

• Expanding monitoring of fatigue causes, behaviors or consequences, including near

misses.

Emphasis was clearly given in the proper compliance to the regulations while it

suggested four points of the above point to be the most important for reducing fatigue:

• proper implementation of the ISM-Code

• optimizing the organization of work on board vessels

• lengthening of the rest period

• reducing administrative tasks on board vessels.

4.4.6 NTUA’S IDENTIFYING AND ASSESSING NON-TECHNICAL SKILLS

ON GREEK MARITIME OFFICERS (2010)

In this survey the NTUA‘s School of Naval Architecture and Marine Engineering2

presented the findings from an attitude questionnaire[110], conducted in captains

A&B and 1st & 2nd engineering officers of Greek owning ship, that cover various

aspects of the relation between human element and maritime safety. In this research

project, Crew Resource Management (CRM) skills analysis was introduced in

maritime industry. This term stands for the critical cognitive, social, and interpersonal

resource skills that complement technical abilities so as to contribute a safe and

efficient task performance. In this outline the basic categories of non-technical skills

in safety critical industries include situation awareness, decision-making,

communication, teamwork, leadership, managing stress, coping with fatigue.

The questionnaire included questions about fatigue perception and hours of sleep

which provided interesting findings regarding fatigue issues in the maritime domain.

The most important findings included:

Officers of the deck stated to sleep less uninterrupted hours, but more in total

number than their shipmates engineering officers

The lack of sleep on officers of the deck of Ro-Ro ships where the scheduled

departures – arrivals, transits through restrained waters, require all officers of

the deck on task, which consequently allow even less available hours for sleep.

STCW 95 followed, even though marginally, about total hours of sleep in

every 24 hour

2 Laboratory of Maritime Transport



STCW 95 is not followed by the hours of rest in a non-routine day where

transit through restrained waters, loading – unloading or emergency situations

took place.

The statistical data collected by this survey served as our main source for assigning

the prior probabilities distribution in our Bayesian Network.

4.4.7 CREW ENDURANCE MANAGEMENT SYSTEM

The Crew Endurance Management System [103] was assembled by the US Coast

Guard in order to help maritime companies ensure the highest possible levels of

performance and safety. The guide focused in two parts:

The identification of specific factors affecting crew endurance in their

particular operations

The management of these factors toward optimizing crewmember endurance

CEMS also addressed the problem of Circadian Rhythm disruption (The Red Zone)

pinpointing that during these drowsy periods the decrease of alertness levels may

cause human errors. Furthermore, the program proposed three steps for the proper

implementation of CEMS in maritime operations, as listed below:

Phase I: Program development which consists of 4 parts:

1. Receive Training in CEM Principles and Practices

2. Identify System Areas of Risk

3. Identify Specific Risk Factors

4. Propose Modifications

Phase II: Program Deployment

Phase III: Program Assessment

Figure 4. 12: Factors that affect the Marine's endurance (CEM) [103]



At this point it is essential to highlight the individual choices node proposed by CEM

system. Fatigue may be the result of a variety of different factors but the most

underrated in literature is the personal judgment and decisions of each individual that

lead to its prevalence. In conclusion, it is essential for an individual to have proper

knowledge about fatigue in order not to underestimate the consequences of his actions

and choices.

Figure 4. 13: Factors that affect Fatigue according to Crew Endurance Management

[103]

4.4.8 FATIGUE AT SEA - A FIELD STUDY IN SWEDISH SHIPPING

This study [52], performed by VTI, collected data about the fatigue level of bridge

watch keepers in order to revise earlier sleep models, and devise innovative solutions

for the shipping industry.

Data collection included interviews with shipping companies and a field study

onboard 13 cargo vessels. 32 participants took part in representing two watch

systems; 2-watch and 3-watch. Subjective sleepiness and stress estimations were

performed once every hour. EOG was used to record eye movement behavior.

Reaction time test was made to examine performance. 3-watch participants are more

satisfied with their working hours and working situation. Tendencies indicate that 2-

watch participants are a bit more tired, whereas the stress is the same. All are less

sleepy and less stressed at home. Time on shift had effect on sleepiness. The highest

KSS (Karolinska Sleepiness Scale) scores were recorded in the late night and early

morning. After night shift the reaction times have higher variance and more long



reaction times are present. The mean value after night shift was significantly higher

than after day shift.

Fatigue measurement methods included the KSS, questionnaires, interviews and

reaction time tests provided the following results:

Figure 4. 14: The Karolinska Sleepiness Scale [52]

Most Swedish companies do not regard fatigue as an important concern

The majority of participants stated that 8 hour sleep is the minimum for

avoiding fatigue issues

Fatigue monitor equipment would be welcomed by the majority of personnel

Fatigue is affected by the shift/watch system (shift rotation)

The average amount of sleep in 24 hours is between 6-7 hours, regardless the

shift system

Sleep quality is low (disturbed)

4.5 LEGISLATION ON FATIGUE IN THE MARITIME SECTOR

4.5.1 IMO & THE ISM CODE

The IMO recognized that there is a direct connection between health and safety of a

ship‘s crew and the way that the vessel is managed. In an attempt to make sure that

shipping companies and operators take a closer look on the potential impact of their

management strategies, the IMO established the International Safety Management

Code.

International Maritime Organization's International Code, International Management

for Safe Operation of Ships and for Pollution Prevention (the ―ISM Code‖) was

adopted in November 1993. During Phase I the ISM Code required the establishment



of a certified safety management system July 1, 1998. The main goals of the Code

were to promote safety at sea, prevent injuries and loss of life and finally setting rules

to diminish potential damage to the environment and to property. In addition, it also

required from shipping firms to establish specific safety objectives, based on their

sphere of operations, and this includes objectives related to crew fitness-for-duty

(which incorporated fatigue factors) and control of human error.

Phase II of the ISM Code became mandatory on the first of July, 2002 and is to be

applicable to all vessels above 500 gross tonnage which were not covered under Phase

I. Phase II introduces a new safety management system, a system that prescribes the

management procedures for safety and pollution prevention for ships or offshore

structures. It has also set forth policy and practice relating to factors that directly or

indirectly affect crewmembers, their roles and responsibilities, work hours, reporting

schemes, paperwork requirements, etc. The responsible parties were required to

provide the resources and support needed to implement and maintain the safety

management system. The Code also requires certification of compliance with the

requirements of the ISM Code and the safety management system [58].

4.5.2 IMO AND STCW

A different aspect that relates to ship management concerns manning issues. In 1995,

the IMO made new amendments to the international Standards for Training,

Certification, and Watchkeeping (STCW) Convention. These proposed regulations

have played an important part in improving the mariners‘ readiness and capabilities.

Some of the more important provisions of the 1995 amendment include:

• Requires certification of masters, officers, and ratings only when they meet the

requirements for service, age, medical fitness, training, qualification, and passing

examinations

• Requires special training for certain types of ships (tankers, Ro-Ro, and passenger

ships)

• Requires Flag States establish and enforce rest periods for watchkeeping personnel.

Watchstanding personnel must be provided a minimum of ten hours of rest for every

24-hour period (the ten hours may be divided into two periods, but one must be of at

least six uninterrupted hours)

• Requires that watch systems be arranged so that watchkeeping personnel is not

impaired by fatigue

• Requires instructors and assessors be qualified for the types of training or

assessment of competence of seafarers. Those involved in training and/or assessment



must be qualified in the task for which the training/assessment is being conducted

[58].

STCW is the most important instrument of the marine industry that can affect the life

on board. Through STCW fatigue can be managed by established limits on work

hours and minimum rest periods for seafarers. Major revisions to the International

Convention on Standards of Training, Certification and Watchkeeping for Seafarers

(the STCW Convention) and its associated Code have been adopted at a Diplomatic

Conference in Manila, thereby ensuring that the necessary global standards will be

prepared to train and certify seafarers to operate technologically advanced ships for

some time to come. The Conference was held in Manila from 21 to 25 June under the

auspices of the International Maritime Organization (IMO). Amendments adopted in

2010 will be implemented on 1 Jan 2012. These amendments addressed fatigue as a

major issue in the maritime industry and proposed revised requirements on work

hours and rest and new requirements for the prevention of drug and alcohol abuse, as

well as updated standards relating to medical fitness standards for seafarers.

4.5.3 ILO 180

The International Labor Organization has also participated in the attempts to reduce

fatigue at sea. The ILO 180, Convention concerning seafarers' hours of work and the

manning of ships, (1996), set limits relating to non-watch-keepers work hours. The

main purpose of these regulations was to mitigate fatigue in marine workplaces (e.g.

engine rooms).

The most important achievement of ILO 180 was setting specific work hours

limitations for seafarers that are listed below:

Maximum hours of work shall not exceed:

(i) 14 hours in any 24-hour period and

(ii) 72 hours in any seven-day period or

Minimum hours of rest shall not be less than:

(i) ten hours in any 24-hour period and

(ii) 77 hours in any seven-day period.



CHAPTER 5 INTRODUCTION TO BAYESIAN

NETWORKS



5 INTRODUCTION TO BAYESIAN NETWORKS

This chapter is a brief introduction to probabilistic graphic modeling and Bayesian

Networks. It must be stated that belief networks are far more complex and demanding

than presented in this thesis. The lack of space and the need for a rather synoptic

presentation has left important elements of Bayesian Networks unaddressed. For

deeper understanding of the capabilities and potential of BN the author suggests the

publications of J.Pearl (1988) and F.Jensen (1996 & 2007), which are listed in the

references [73] [40] [41], regarding intelligent systems and BN. Most part of this

chapter was based to the thorough contents of these books.

5.1 AN INTRODUCTION TO UNCERTAINTY AND GRAPHICAL MODELS

5.1.1 UNCERTAINTY

D. Hubbard has described uncertainty as the lack of certainty, a state of having limited

knowledge in which it is impossible to describe precisely existing state or future

outcome. Decision making is widely recognized by engineers as an integral part of the

whole engineering design process. Because of the fact that uncertainty has a

significant impact on decision making, the engineering community has tried to

manage uncertainty via innovative methods and intelligent systems. The presence of

uncertainty in the scientific and professional world has led scholars to develop tools in

order to address issues such as imprecision and event dependence. The most reliable

tool for modeling uncertainty has been the use of probabilities theory.

Jensen (1996) [40] proposes a way to incorporate uncertainty in a rule based system

by extending the production rules to the following format.

If condition with certainty x then fact with certainty f(x), where f is a function

The inference system must be extended with new inference rules, which shall ensure a

coherent reasoning under uncertainty. For example, if x is the certainty of inferring C

from A and y the certainty via inferring C from B, we can conclude that the certainty

of C is the inference of the function g(x , y).

The answer to uncertainty was given by decision theory experts through modeling

classical probability theory into a more precise mathematical context. Through these

methods, researchers managed to quantify reasoning under uncertainty by taking

advantage of the capabilities of modern computer systems. One of the most effective

tools to manage and represent uncertainty has been the use of probabilistic graphical

models, a modeling method that will be discussed in the following paragraph.

Other methods to deal with uncertainty include:

• Three-valued logic: True / False / Maybe



• Fuzzy logic (truth values between 0 and 1)

• Non-monotonic reasoning (especially focused on Penguin informatics)

• Dempster-Shafer theory (and an extension known as quasi-Bayesian theory)

• Possibabilistic Logic [63].

5.1.2 PROBABILISTIC GRAPHICAL MODELS

A graphical model is a probabilistic model for which a graph denotes the conditional

independence structure between random variables. Probability theory, statistics and

machine learning are some of the fields that have benefitted from the use of these

models. Many scientists describe them as a marriage between probability and graph

theory. The three fundamental cornerstones of graphical models are representation,

inference, and learning.

Michael Jordan (1998) describes the primary purpose and function of graphical

models:

„An integral part in the idea a graphical model is the notion of modularity -- a

complex system is built by combining simpler parts. Probability theory provides the

glue whereby the parts are combined, ensuring that the system as a whole is

consistent, and providing ways to interface models to data. The graph theoretic side

of graphical models provides both an intuitively appealing interface by which humans

can model highly-interacting sets of variables as well as a data structure that lends

itself naturally to the design of efficient general-purpose algorithms. Many of the

classical multivariate probabilistic systems studied in fields such as statistics, systems

engineering, information theory, pattern recognition and statistical mechanics are

special cases of the general graphical model formalism – examples include mixture

models, factor analysis, hidden Markov models, Kalman filters and Ising models. The

graphical model framework provides a way to view all of these systems as instances

of a common underlying formalism. This view has many advantages -- in particular,

specialized techniques that have been developed in one field can be transferred

between research communities and exploited more widely. Moreover, the graphical

model formalism provides a natural framework for the design of new systems‟.3

5.2 BAYESIAN NETWORKS AND BAYESIAN INFERENCE

5.2.1 WHAT IS A BAYES NET?

One of the most prevalent and effective graphical models to manage uncertainty is the

Bayesian Network which was popularized in 1988 by Judea Pearl in Probabilistic

Reasoning in Intelligent Systems. A Bayesian network, belief network or directed

acyclic graphical model, is a probabilistic graphical model that correlates the

conditional dependencies of a number of random variables with the use of a Directed

3 M. I. Jordan . "Learning in Graphical Models". MIT Press. 1998

http://en.wikipedia.org/wiki/Probabilistic_model

http://en.wikipedia.org/wiki/Graph_%28mathematics%29

http://en.wikipedia.org/wiki/Conditional_independence

http://en.wikipedia.org/wiki/Conditional_independence

http://en.wikipedia.org/wiki/Random_variable

http://en.wikipedia.org/wiki/Graphical_model



Acyclic Graph. A directed acyclic graph is a directed graph with no directed cycles.

The formation of a DAG includes vertices and directed edges, each edge connecting

one vertex to another so that a cyclic route is impossible to appear. In the following

figure we can see the fundamental difference of an acyclic and a non-acyclic graph.

Figure 5. 1: The difference between non-acyclic (left) and acyclic (right) graphs [40]

A common example of a Bayesian network is the representation of the probabilistic

relationships between diseases and symptoms. Given symptoms, the network can be

used to compute the probabilities of the presence of various diseases.

The nodes in Bayesian Networks represent random variables in the Bayesian sense:

they may be observable quantities, latent variables, unknown parameters or

hypotheses. Edges represent conditional dependencies between parent and child

nodes. Each node is associated with a probability function that takes as input a

particular set of values for the node's parent variables and gives the probability of the

variable represented by the node. These probabilities of the child nodes are sorted in

Conditional Probabilities Tables or CPT‘s as it can been seen in the following figure.

Fever and spots are the parent nodes that affect in terms of conditional probabilities

the child node Measles. The CPT shows for example what is the probability of

Measles P(M) when Fever state is T and Spots state is T (figure 5.2).

http://en.wikipedia.org/wiki/Directed_graph

http://en.wikipedia.org/wiki/Cycle_graph#Directed_cycle_graph

http://en.wikipedia.org/wiki/Vertex_%28graph_theory%29

http://en.wikipedia.org/wiki/Edge_%28graph_theory%29

http://en.wikipedia.org/wiki/Random_variables

http://en.wikipedia.org/wiki/Bayesian_probability

http://en.wikipedia.org/wiki/Latent_variable

http://en.wikipedia.org/wiki/Probability_function

http://en.wikipedia.org/wiki/Glossary_of_graph_theory#Directed_acyclic_graphs



Figure 5. 2: A casual network and its Conditional Probabilities Table

5.2.2 CONDITIONAL PROBABILITIES AND INDEPENDENCE IN

BAYESIAN NETWORKS

5.2.2.1 CONDITIONAL PROBABILITIES AND BAYES THEOREMS

In the previous paragraph we mentioned the term conditional probability. The basic

concept in the Bayesian treatment of certainties in causal networks is conditional

probability . Whenever a statement of the probability, P (A), of an event A is given,

then it is given conditioned by other known factors. A statement like ―the probability

of the die turning up 6 is 1/6‖ usually has the unsaid prerequisite that it is a fair die or

rather, as long as I know nothing of it, I assume it to be a fair die. That means that

the statement should be ―Given that it is a fair die, the probability ....‖ In this way,

any statement on probabilities is a statement conditioned on what else is known.

A Conditional probability statement is of the following kind:

“ Gi ven t h e even t B , t h e pr ob ab i l i t y o f t h e even t A i s x .”

The notation for the statement above is



P (A |B ) = x .

It should be stressed that P(A|Β ) = x does not mean that whenever Β is true then the

probability for A is x . It means that if Β is true, and eve r y t h in g e l se kn ow n i s

i r r e l evan t f o r A , then P( A ) =x .

The fundamental rule for probability calculus is the following:

P (A /B) P (B) =P (A, B) (5.1)

Where P(A/B) is the probability of the joint event A ^ B. Remembering that

probabilities should always be conditioned by a context C, the formula should read

( / , ) ( / ) ( , / )P A B C P B C P A B C (5.2)

From (5.1) it follows that P ( A/ B)P ( B) = P (B / A )P ( A) and this yields the well

known B a yes r u l e :

P (B / A )= P (A / B )P (B ) / P (A ) (5.3)

Bayes' rule conditioned on C reads

( / , ) ( , )

( / , )( / )

P A B C P B CP B A C

P A C (5.4)

[40]

Joint probability is the probability of two events in conjunction. That is the probability

of both events together. The joint probability of A and B is written P(A^Β), P(AB) or

P(A,B).

Marginal probability is then the unconditional probability P(A) of the event A; that is,

the probability of A, regardless of whether event B did or did not occur. If B can be

thought of as the event of a random variable X having a given outcome, the marginal

probability of A can be obtained by summing (or integrating, more generally) the joint

probabilities over all outcomes for X. For example, if there are two possible outcomes

for X with corresponding events B and B', this means that.

( ) ( ) ( ')P A P A B P A B . This is called marginalization.

5.2.2.2 D- SEPERATION AND CONDITIONAL INDEPENDENCE

Types of connections in Bayesian Networks

Bayesian Networks are casual networks that propagate certainty/evidence through 3

different ways as discussed below:

http://en.wikipedia.org/wiki/Joint_probability

http://en.wikipedia.org/wiki/Marginal_probability

http://en.wikipedia.org/wiki/Random_variable

http://en.wikipedia.org/wiki/Summation

http://en.wikipedia.org/wiki/Integral



A) Serial connection

Figure 5. 3: Serial Connection [40]

A has influence on B which in turn has influence on C. When there is evidence given

about B, the communication between A and C is blocked.

B) Diverging connections

Figure 5. 4: Diverging Connection [40]

In this connection A is the parent node while B, C and X are the child nodes. If

evidence is given about A, communication between child nodes is blocked.

C) Converging connections

Figure 5. 5: Converging Connection [40]



In this connection A is the child node while B,C and X are the parents. If evidence is

given about A, communication is opened between its parents.

D-separation

These three types of connections cover all the ways in which certainty can be

transmitted through a variable, and following the rules it is possible to decide for any

pair of variables in a causal network whether they are dependent, given the evidence

entered into the network. The rules are formulated in the following, (Jensen 1996):

Definition (d-separation)

―Two variables A and B in a causal network are d - s epa ra t ed if for all paths

between A and Β there is an intermediate variable V such that either

- The connection is serial or diverging and the state of V is known

Or

- The connection is converging and neither V nor any of V s descendants have

received evidence.

If A and Β are not d-separated we call them d - co n nec ted . ”

Conditional independence

D-separation, the blocking of the transmission of certainty through a casual network

is, in the Bayesian calculus, reflected in the concept of c o n d i t i on a l

i nd ep en den ce . T h e variables A and C are i n d ep end en t g i ven the

va r i ab l e Β if:

( / ) ( / , )P A B P A B C (5.6)

In plain words, it means that whether there is evidence about B, no knowledge

regarding A will change the probability of C [40] .

5.2.2.3 BAYESIAN NETWORK

According to Jensen (1996), a Bayesian Network consists of the following:

A set of va r ia b l es and a set of d i r ec t ed ed ges between variables.

Each variable has a finite set of mutually exclusive states.

The variables together with the directed edges form a d i r ec t ed

a cyc l i c graph (DAG)



To each variable A with parents B (1 ) . . . , B (n ) there is attached a

conditional probability table P(A | Β (1 ) , … B (n ) .

A Bayesian Network can best be described by the functioning of the chain rule:

Let X be a Bayesian network over U == {Αη,….Am}. Then the joint probability

distribution P(U) is the product of all conditional probabilities specified in X:

( ) ( | ( ))P U P Ai pa Ai

Where pa(Ai) is the parent of Ai

5.2.3 BAYESIAN INFERENCE

The most common problem we wish to solve using Bayesian networks is probabilistic

inference. Bayesian inference is a method of statistical inference in which some kind

of evidence or observations are used to calculate the probability that a hypothesis may

be true, or else to update its previously-calculated probability. The easier way to

understand the meaning of the Bayesian Inference is through the Bayes Theorem.

( / ) ( )

( / )( )

P A B P BP B A

P A

B represents a specific hypothesis, which may or may not be some null

hypothesis.

A represents the evidence that has been observed.

P(B) is the prior probability of H that was inferred before new evidence

became available.

P(B/A) is the conditional probability of seeing the evidence E if the

hypothesis H happens to be true. It is also called a likelihood function when it

is considered as a function of B for fixed A.

P(A) is the marginal probability of A: the a priori probability of witnessing

the new evidence A under all possible hypotheses. It can be calculated as the

sum of the product of all probabilities of any complete set of mutually

exclusive hypotheses and corresponding conditional probabilities.

P(B/A) is the posterior probability of B given A and is the new estimate of the

probability that the hypothesis B is true, taking the evidence A into account.

The factor P(A/B) / P(A) represents the impact that the evidence has on the belief in

the hypothesis. If it is likely that the evidence A would be observed when the

hypothesis under consideration is true, but when no hypothesis is assumed, it is

inherently unlikely that A would have been the outcome of the observation, then this

factor will be large. Multiplying the prior probability of the hypothesis by this factor

would result in a larger posterior probability of the hypothesis given the evidence.

Conversely, if it is unlikely that the evidence A would be observed if the hypothesis

http://en.wikipedia.org/wiki/Statistical_inference

http://en.wikipedia.org/wiki/Probability

http://en.wikipedia.org/wiki/Null_hypothesis

http://en.wikipedia.org/wiki/Null_hypothesis

http://en.wikipedia.org/wiki/Prior_probability

http://en.wikipedia.org/wiki/Conditional_probability

http://en.wikipedia.org/wiki/Likelihood_function

http://en.wikipedia.org/wiki/Marginal_probability

http://en.wikipedia.org/wiki/Posterior_probability



under consideration is true, but a priori likely that A would be observed, and then the

factor would reduce the posterior probability for B. Under Bayesian inference, Bayes

theorem therefore measures how much new evidence should alter a belief in a

hypothesis.

Later in this chapter, we will discuss a specific example in order to show the use of

Bayesian Networks and the importance of the Bayesian Inference in practice.

5.3 PRESENTATION OF A BAYESIAN NETWORK

5.3.1 TYPICAL PRESANTATION OF A PROBLEM OF UNCERTAINTY

Mr. Holmes lives in Los Angeles. One morning when Holmes leaves his house, finds

that the grass in his yard is wet. The grass could be wet because of rain (R), or

because he has forgotten to switch off the automatic watering system (S). His faith of

these two potential events increases.

Holmes also finds that the neighbor's lawn, Dr. Watson, is also wet. Now Holmes is

almost sure that it has been raining. A representation of the situation is shown in

Figure 5.6, where the rain (R) and the automatic watering system (S) are causes for

the wet grass of Holmes. Only the rain can cause Watson grass to be wet.

Figure 5. 6: Graphical presentation of the 'wet grass' example [40]

When Holmes realizes that his grass is wet, apply reasoning in the opposite direction

from that of explanatory arrows. Since both incidence functions that result in H

increase, certainty for the R and S also increases. The increased certainty of R in turn

creates a greater certainty of W.



Consequently, when Holmes inspects the lawn of Watson and discovers that it is also

wet, he rapidly increases the certainty of R. The next step in the reasoning is difficult

for machines, but natural for people and in-depth explanation is required: the wet

grass of Holmes explained, and so there is now no reason to think that the automatic

watering system was open. Therefore, the certainty of S is reduced to the initial value.

The in-depth explanation is yet another example of dependence changing on the

available information. In its original condition, when nothing is known, the R and S

are independent. However, when we have information related to the lawn of Holmes,

then the R and S become dependent.

5.3.2 USE OF BAYESIAN NETWORKS IN A TYPICAL EXAMPLE

In this paragraph we will transform the problem of 5.3.1 into a Bayesian Network,

while explaining the practical use of the graphical model and apposing all the

calculations that take place.

Assume that the prior probabilities for the R and S nodes are P(R) = (0.2,0.8), and

P(S) =(0.1,0.9). The remaining probabilities are shown in Table 5.1. Initially, we

calculate the root chances for W and H by the formula 5.1. That is, first calculate the

P (W, R) and then marginalize R. The result is P (W) = (0.36, 0.64).

Table 5. 1: P (W/R) and P (H/R, S)

The calculation of P (H, R, S) follows the same procedure, except that the product in

this case is:

P (H, R, S) = P (H / R, S) P (R, S)

Since the R and S are independent we have:

P (H, R, S) = P (H / R, S) P (R) P (S)

The result is given in Table 5.2 Marginalizing R, S out of P(H,R,S) gives P (H) =

(0.272 , 0.728). Now we have constructed joint probability tables for two of the

clusters, (W, R) and (H, R, S), with variable R common.



Table 5. 2: Prior probability table for P (H,R,S)

The certainty H=y is used for updating P(H,R,S) by deleting all entries with H=n and

dividing by P(H=y). The result will be a probability table with all entries summing to

one, therefore we do not need to calculate P(H). After all the entries with H=n have

been erased (Table 5.3.), we simply normalize the table by dividing the sum of the

remaining entries (see Table 5.4.).

Table 5. 3: P (H,R,S) with entries H= n eliminated

Table 5. 4: Calculation of P*( H,R,S) = P(H,R,S/ H=y)

We then get:

P*(R=y) = 0.736 and P*(S=y)= 0.339 .

Use the P*(R) for updating P(W,R) (see Table 5.5)

P*(W,R)= R(W/R)*P*(R) =P(W,R) P*(R)/ P(R)



Table 5. 5: Calculation of P*(W,R)= P(W/R)* P*(R) = P(W,R) P*(R) / P(R)

Now we use the fact that W=y for updating the distribution for (W, R) (see Table 5.6).

We take P**(R=y) = 0.93.

We still have to calculate the P**(S) = P(S/W=y, H=y). The result must reflect the

impact of the in – depth explanation; since the wetted grass is explained by the rain,

the probability for S=y should be reduced to its initial value.

The calculation shall follow the same pattern. A message on to P**(R) sent by (W,R)

to (H,R,S) (see Figure graph)

P** (H,R,S) = P*(H,R,S) P**(R) /P*(R)

Using marginalization we get P**(S=y)= 0.161 .

Table 5. 6: P**(W,R) =P( W,R /W=y, H=y)

Table 5. 7: P**(R,S) =P( R,S /H=y,W=y)

The reason for the probability of S not dropping to its initial value of 0.1 is that

Watson may have forgotten the system open as well. This is reflected in the

probability P (W=y/R=n) =0.2.



5.3.3 USING GENIE

GeNie was developed by the University of Pittsburgh as an instrument to manage

Graphical Models in decision making. It is characterized by a user friendly interface

and a plethora of different functions within the Bayesian Networks domain. In this

paragraph we will present a simple demonstration of using Genie for the example of

5.3.1 and 5.3.2.

Step 1

Create a new network by clicking on the FILE NEW NETWORK

Step 2

Create the nods for this network by clicking on the node icon in the toolbar. For this

particular network the nodes are H(HOLMES), W(WATSON), R(RAIN),

SYSTEM(S).

If you set the nodes right you have completed the first part of your BN.



Step 4

Define the dependencies between the nodes. For example R has impact on W and H.

Select the arc icon and drag the arrow from the parent node to the child node.

Step 5

Set the states for each node. In this part, double click on the target node and then on

DEFINITION. There you can name each state and set the prior probabilities for the



parent nodes. For example on the R node the states are yes and no and the

probabilities 0.2 and 0.8 respectively.

Step 6

Set the conditional probabilities tables for the child nodes. The following table is a

2x4 matrix which includes all the possible scenarios and their outcomes in

probabilistic terms. Note that when there is more than one parent nodes conditional

independence must be present.

Step 7



Setting evidence: Since we are certain that H=y we left click on the node and chose

SET EVIDENCE and set it to yes. We then follow the same procedure for W=yes.

Step 8

Updating the network and getting the Bayesian Inference. In order that information is

propagated throughout the network we click on the update icon (lightning). Then

when all nodes and tables are updates we can see the posterior probabilities of its

node by dragging the mouse on the tick icon in the bottom right side of the node.

We notice that R= y is 0.933 which is the exact same value we get after marginalizing

Table 5.7 in the previous paragraph.



5.3.4 NOISY OR GATES

The most integral part of a Bayesian network is the setting of the CPT‘s. While prior

probabilities are generally easy to acquire, the dependence of the parents-child nodes

is a complex procedure especially when the number of parents is high. In the example

cited in 5.3.1 the parent nodes to H were S and R creating a 2x4 CPT. If for example

the parent nodes were 8 and the states of H were 4 we would have a 4x8 CPT making

the conditional probabilities assignment extremely difficult.

The noisy-or principle is a modeling trick that help us overcome these difficulties.

Noisy-OR gates are usually used to describe the interaction between n causes X1,

X2….Xn and their common effect Y. The causes Xi are each assumed to be sufficient

to cause Y in absence of other causes and their ability to cause Y is assumed

independent of the presence of other causes [42].

The Noisy OR gate applies when there are several possible causes X1, X2, . . . , Xn of

an effect variable Y, where (1) each of the causes I has a probability P(i) of being

sufficient to produce the effect in the absence of all other causes, and (2) the ability of

each cause being sufficient is independent of the presence of other causes. The above

two assumptions allow us to specify the entire conditional probability distribution

with only n parameters P(1, 2, . . . , n). P(i) represents the probability that the effect Y

will be true if the cause X is present and all other causes Xj, j/=i, are absent [69].

Jensen 1996 describes the Noisy or gate in the following excerpt, [40] (pg 48-51):

Let A 1. . . . . . , An „ be binary variables listing all the causes of the binary variable

B . Each event Αi· = y causes Β = y unless an inhibitor prevents it, and the

probability for that is q i (see Figure 5.7).

That is, Ρ (Β = n | Α = y) = q i . We assume that a l l i n h i b i t or s a re

i nd ep en den t . Then P(B = n /A 1 , A 2 . . . ,A n) = Π j € Y q j where Y is the set of

indices for variables in the state y. For example:

P (B = y / A1 = y , A2 = y , A3 = ……. = A n = n )

= 1 -P (B = n /A 1 =y , A2 =y ,A 3 =y… . = A n =n )

= 1 – q1 · q2

Figure 5. 7: Noisy OR Gates example



We require P (B = y / A 1 =…… = A n = n ) to be zero. This may seem

to restrict the applicability of the approach. However, as in the example

above, if P (B = y ) > 0 when none of the causing events in the model are

on, then introduce a background event which is always on.

The complementary construction to noisy or is called n o i s y an d . A set of

causes shall all be on in order to have an effect. However, the causes have

random inhibitors which are mutually independent.

The noisy or-gate can be modeled directly without performing the calcu-

lations. This highlights the assumptions behind the noisy or-gate. If a cause

is on, then its effect may be prevented by an inhibitor, and the probabilities

for the inhibitors to be present are independent.

5.4 COMMON BAYESIAN NETWORK APPLICATIONS

Some of the fields that have benefited from the use of Bayesian networks are:

Medicine

Biology

Law

Engineering

Computer science

Artificial Intelligence

5.4.1 GENERAL

In this paragraph we are going to present some common example of Bayesian

Networks applications in various scientific domains [108].

A) ENVIRONMENT

Assessing the viability of fish populations (Lee and Reiman 1997)

Management of the food chain of lakes (Varis et al. 1990)

Eco-risk assessment (Pollino and Hart 2005)

Farming systems modeling (Varis anf Kuikka 1997)



B) BIOLOGY

Brain function studies, (Rajapakse et al, 2007)

Protein structure, (Bradford,Needham et al., 2006)

Genes networks analysis, (Mittal et al., 2005)

Non-Parametric regression in genetics, (Imoto et al., 2002)

B) MEDICINE

Risk Analysis and Management Methodology (CRAMM), (Maglogiannis,

Zafiropoulos et al., 2005)

Prognosis Bayesian networks, (Verduijin et al., 2007)

Cancer diagnosis (Nicandro Cruz-Ramirez et al, 2007)

C) INDUSTRY

Jet engines failure (Ferat et al., 2007)

Supply chain management (Han-Ying Kao et al., 2005)

Oil products impact (Cattanach et al., 1995)

D) ECONOMICS AND OPERATIONS RESEARCH

Corporate alliances analysis, (Letterie et al., 2006)

Informations recovery via conversation (Kyoung-Min Kim et al., 2006)

Robotics, (Lazkano et al., 2006)

F) EDUCATION

Internet based learning (Garcia et al., 2005)

Digital image libraries (Kahn Jr. 2001)

G) RISK MANAGEMENT AND ASSESMENT

Corporations risk and decision making (Bonafede and Giudici, 2007)

Risk distribution databases, (Druzdzel et al., 1995)



5.4.2 MARITIME

There are several studies within the Maritime domain that used Bayesian Network in

order to treat uncertainty.

Trucco, Cagno et al (2007) [102] attempted to include Human and Organizational

Factors (HOF) in their analysis of risk. Their approach was developed to address the

sector of maritime industry, but it can be used in other sectors, as well. A Bayesian

Network was created in order to model the Maritime Transport System (MTS), taking

into consideration the various factors that are involved as well as their reciprocal

interdependence. The particular model was used to quantify the importance of HOF in

the risk analysis during the fundamental stages of designing a High Speed Craft

(HSC). The study focused on the open sea collision applied via an innovative method

of unification of the Fault Trees Analysis (FTA) and Bayesian Belief Network

regarding the influences of organizational operations and regulations, as they are

addressed by the IMO and the FSA. This approach allowed the recognition of cross-

correlation of probabilities between the basic events in the case of a collision as well

as the network of faith that includes the operational and organizational factors.

In their study Norrington et al. (2007) presented a methodology of reliability

modeling regarding research and operations of rescue of coast guard coordinative

centers. Various sources are found by governmental reports, interviews or supervision

personnel. The incorporation of information was achieved with the use of belief

networks [108].

Bayesian Networks were recently used by the Norwegian FSA for estimating the

safety of passenger ships (2005). This particular study focused in finding effective

Risk Control Options (RCOs) in the sector of marine navigation. In order to reduce

the frequency of collisions and groundings, the study indicates that following RCOs

provide considerably improved safety of marine navigation with a limited cost:

Electronic Chart Display and Information Systems (ECDIS)

The track control system (track control)

The Automatic Identification System (AIS) combined with the operational

radar

The improved planning of bridges

The improved learning of navigator [108].

5.4.3 FATIGUE

Bayesian networks are considered by many scholars as a state of the art instrument to

deal with uncertainty. Fatigue, which is the main subject of this thesis, is an issue that

http://www.google.gr/url?sa=t&source=web&cd=5&ved=0CDMQFjAE&url=http%3A%2F%2Fen.wikipedia.org%2Fwiki%2FAutomatic_Identification_System&ei=W3F2TIaRNuqg4Qbpm9m5Bg&usg=AFQjCNG_MzcCvWdVoWvDUR5JGFJboQPCgg



has to manage uncertainty in order to predict its prevalence and potential outcomes.

Belief networks give researchers this particular versatility.

In their study Peilin, Qiang and Looney [42] [43] introduce a probabilistic model

based on Bayesian Networks that gives the probability of human fatigue by acquiring

information from various visual cues and certain relevant contextual information. A

brief review regarding the physiological and psychological aspects of fatigue is made

in order to pinpoint the risk factors that create it. The influence of these factors, that

were acquired via quantification of data from various relevant fatigue studies, were

the information variable in their model. Specific emphasis was given to visual

characteristics such as eye-lid movement, gaze, head movement, and facial

expression. These signs were recorded as the main signal for inferring fatigue. The

most difficult part of this endeavor was the parameterization of the model (CPT‘s)

which was made possible by the extraction of data from related studies and subjective

knowledge as well. The use of the noisy or gates played a significant part in

eliminating complexity problems of joint distributions. The inference results produced

by running the fatigue model using Microsoft BNs engine MSBNX ,a program very

similar to GeNie.

While their model managed to produce a static representation of fatigue, it failed to

capture the dynamic dependency between fatigue and time. In order to overcome this

limitation the static model was reproduced into a Dynamic Bayesian Network which

can record the impact of time. By connecting various visual fatigue evidences with

time, this dynamic fatigue model leaded to a more robust and accurate estimation of

fatigue.

This particular model and many statistical data that came with it served as the basis of

our own static model for fatigue probability inference aboard ships. Peilin, Qiang and

Looney used [43] a computer vision system in order to record visual signs in different

time periods, a practice which allowed them to give their model a dynamic

perspective. Unfortunately, monitoring fatigue on board is a more complex task that

cannot be recorded by any available computer system. Most of our sources and

evidences came from relevant research regarding the issue as well as seafarer‘s diaries

and interviews.

5.5 WHY BAYESIAN NETWORKS ARE SO HELPFUL TO THESIS THESIS.

ADVANTAGES DISADVANTAGES

In this paragraph, we discuss the advantages of using Bayesian networks and why

they are the most effective means to predict fatigue in a maritime environment. The

parameterization of a fatigue model may be a very complex and demanding task,

nevertheless, the proper use of statistical data and the capabilities of Bayes theories

can become a very accurate and practical way to reduce fatigue at any workplace.



Bayesian networks provide a powerful instrument to capture the uncertainties,

dependencies and dynamics exhibited by fatigue. Through BN it becomes easier to

incorporate information and observations, model uncertainty and aspects of fatigue

that are difficult to quantify and above all, setting a graphical model with a huge

number of different sources of data. All these benefits along with the contemporary

research regarding probabilistic models (especially BN software) make Bayesian

Networks the ideal instrument for fatigue inference and a genuine ‗weapon‘ to

overcome any potential limitations.

Moreover, BN can be used, even in the case of missing data in order to learn the

causal relationships and gain an understanding of the various problem domains and to

predict future events [63].

Furthermore, large conditional probabilities tables may require the need of a tool for

extensive calculation and analysis. The popularization of BN has led many

programmers to develop user friendly software as GeNie so these limitations could be

easily put aside.

Petri Myllymäki summarizes some of the BN‘s distinctive advantages over alternative

modeling approaches [FF]:

Decision theory:

Using BN in decision theory for risk analysis can provide the optimal procedure in

decision making.

Consistent, theoretically solid mechanism for processing uncertain information:

BN reduces the probability of an inference to be ambiguous by using consistent

calculus for uncertainty.

Smoothness properties:

Bayesian networks can be easily modified according to the contemporary changes

without having to re-model ‗from scrap‘ through the use of updating.

Flexible applicability:

The same Bayesian network model can be used for solving both discriminative tasks

(classification) and regression problems (configuration problems and prediction).

A theoretical framework for handling expert knowledge:

Expert knowledge can be interpreted into prior probabilities which are independent to

any potential sampling data which allows the interaction of expertise with statistical

data. While using BN expert can finally compare their prior estimation to that of the

data.



A clear semantic interpretation of the model parameters:

They can be used directly without the need of prior learning making them superior to

other forms of networks such as neural networks.

Different variable types:

While most probabilistic models work with a specific type of variables and

distributions, BN overcome that limitations and allow various type of different

variable to coexist in the same domain with no need of transformation.

A theoretical framework for handling missing data:

Marginalization in BN helps us put aside any limitation that comes from missing.

While the main argument against BN is that the parameterization can be very complex

and time consuming, the theoretical context of BN indicates the possibility of creating

a BN with a small number of parameters. This fact has led high tech corporations like

Microsoft and organizations like NASA to develop applications based on Bayesian

Networks for exploiting their numerous capabilities.



CHAPTER 6 A BAYESIAN NETWORK FOR

PREDICTING FATIGUE IN

MARITIME TRANSPORT



6 A BAYESIAN NETWORK FOR PREDICTING FATIGUE IN

MARITIME TRANSPORT

In this chapter we present the nodes of our Bayesian Network regarding fatigue

aboard ships. The categorization was made according to IMO‘s guidelines for fatigue

mitigation. Finally, we set forth our sources, the states of each parent node, the prior

probabilities and the Conditional Probabilities Tables (CPT) that represent the

dependence between the parent and the child nodes. At this point it should be

mentioned that the Noisy OR principle, which was discussed in the previous chapter,

was used in order to calculate the probabilities for the CPT‘s where the parent nodes

were great in number.

6.1 MODEL PRESENTATION

6.1.1 TARGET NODE

1. Fatigue

Fatigue is the target node of this Bayesian Network while the primary goal of its

function is the calculation of the posterior probability. For this node we have tried to

use the same parent-node structure that Ji, Qiang and Looney (2003) [42] introduced

in their paper regarding drivers‘ fatigue. In order to guarantee the overall precision of

our model, we have used certain CPTs of this survey while making specific changes

according to our own subjective judgment and knowledge.

Furthermore, in order to enhance the model‘s reliability the following assumptions

were taken into consideration.

Fatigue is inevitable even if all factors favor its mitigation.

The parent nodes for fatigue are universal for almost every transport domain.

States of fatigue:

Yes

No

The following image is the first of the CPTs that we will provide in order to

demonstrate the connection between the parent nodes and the child one. Those nodes

who are solely parent nodes were given a prior probability distribution that is

presented in similar manner with the CPT‘s

Note: Since we found no better way to present the CPT‘s and prior probability tables,

we provide snapshots from the actual software used. We ask the reader to excuse us

for the quality of these images.



Source: [42] [43]

6.1.2 PARENT NODES

6.1.2.1 CREW SPECIFIC FACTORS

2. Overall Sleep quality

This node represents the overall sleep quality in terms or quantity, environment

condition and the opportunity for occasional naps. In our research regarding fatigue

we have concluded that sleeping quality is by far the most important contributor in

inducing fatigue, a fact that can be easily proved by the results provided by our

network ( see Chapter 7).

States of sleep quality:

Poor

Fair

Source: [42] [90]



Source: [42] [90]

3. Biological clock

The human biological clock, widely known amongst scholars and researchers as

Circadian Rhythms, is a useful instrument for pinpointing the time periods when the

body is awake and the red zone when each individual is vulnerable to the effects of

fatigue and lack of alertness.

States of Biological Clock:

Drowsy

Awake

Source: [42] [14] [103]

4. Physical state

Most surveys and research projects have indicated that fatigue can be the result of

lack of exercise, nutrition issues and aspects of general fitness and well being. For

example, an operator who has no physical exercise routine is significantly more

exposed to fatigue and human error than a person who has taken good care of his

body and general health.

States of Physical state:

Poor



Healthy

Source: [40]

5. Work Condition

The work condition node describes the overall nature of the task at hand. Since the

parent nodes are the type of work in terms of tedium and the heaviness of the

workload, it is made clear what this node actually represents; a very common example

of a poor work condition at sea can result in the helmsman error when a watchkeeper

falls asleep on duty both because of fatigue issues and the monotonous nature of his

task.

States of Work Condition:

Poor

Fair

Sources: [42]

6. Sleep Quantity

Sleep quantity can be described in two different ways:

Hours that an individual has spent sleeping ( in order for all sleeping parts4 to

take place)

Hours that an individual has remained awake.

Heavy workload, low manning levels and emergency situations are some of the

factors that can influence sleep quantity. Sleep deprivation is perhaps the most crucial

part of the overall sleeping quality while staying awake for more than 20 hours can

lead to numerous human errors and general underperformance.

4 Sleep architecture (Chapter 3)



States of Sleep Quantity:

Deprived

Sufficient

Sources: [42] [110] [4] [5]

7. Napping

In chapter 3 we indicated that napping for more than 30 minutes can be beneficial for

seafarers as research has shown that this time is sufficient for being restorative to the

human body and mental performance as well. In accordance to sleep quantity, the

opportunities for naps are connected to the nature of the workload aboard the ship.

States of Napping:

No

Minimum30 min

Sources: [42]

8. Sleep Condition

Sleep condition can best be described as the ability of each individual to sleep by

putting aside distractions, anxiety and other factors. A very common problem for

seafarers, when they have sufficient time for rest, is their difficulty to attain sleep.

The most common distractions for sleeping are stress and anxiety.

States of Sleep Condition:

Bad

Fair



Sources: [110] [90]

9. Circadian intermediate

This intermediate node is used in order to take advantage of [42] CPT regarding jet

lag and time state. Our own research has indicated that both shift rotation and lighting

have huge impact on the Circadian Rhythms so we felt that it was important to

include them in our model as well.

States of Circadian intermediate:

Drowsy

Awake

Sources: [42]

10. Sleep Disorders

A sleep disorder is a medical disorder of an individual‘s sleeping patterns. Some sleep

disorders are serious enough to interfere with normal physical, mental and emotional

functioning. Several surveys have indicated that a great number of mariners suffer

from sleep disorders thus affecting their physical condition, sleep quality and

performance.

Sleep Disorders:

Regular

Rare

Sources: [42] [90] [5]




11. Hours awake

This node is introduced in order to give our model a more dynamic perspective. The

node states are based on the need for hours of sleep for an average adult (8) and

several graphs that (mostly by Sirois, 1998) that pictures the impact of sleep

deprivation and fatigue on alertness. Most probabilities for the CPTs were based on

NTUA‘s survey regarding the average hours of sleep of officers.

States of Hours awake:

12

14

16

18

20

24

More than a day

Sources: [110]

12. Age

While a specific survey [90] has proved that a more experienced is seafarer shows

lower stress levels and greater self confidence than a young inexperienced mariner,

medical research has shown that the overall physical efficiency decreases with age.

Literature has found the age of 45 as the hub of this age-physical condition

relationship.

States of age:

45 <

45 >



Sources: [110]

13. Fitness

General fitness is an essential part for good physical condition and the mitigation of

fatigue. Physical fitness is generally achieved through exercise, proper nutrition and

adequate rest. An overweight operator will get tired more easily than a fit person

while the consumption of drugs or alcohol can also cause fitness and physical

condition problems.

States of Fitness:

Poor

Fair

Sources: [110] [34]

14. Anxiety

Anxiety is a feeling of nervousness, apprehension, fear, or worry. Some fears and

worries are justified, such as worry about a loved one or in anticipation of taking a

quiz, test, or other examination. Problem anxiety interferes with the sufferer's ability

to sleep or otherwise function. Stress is also connected with anxiety while most

mariners identify anxiety as a main contributor for sleep deprivation, bad decision

making, communication problems and underperformance. While it is impossible for a

human to eliminate anxiety completely, we have divided anxiety into normal and high

level. Despite the general scientific approach for anxiety, we have made separate

nodes for stress and anxiety. While many scholars describe the terms as almost

http://en.wikipedia.org/wiki/Physical_exercise

http://en.wikipedia.org/wiki/Nutrition

http://en.wikipedia.org/wiki/Rest

http://www.emedicinehealth.com/script/main/art.asp?articlekey=2279

http://www.emedicinehealth.com/script/main/art.asp?articlekey=59135



synonymous, in our network high stress levels lead to anxiety, which serves as an

intermediate state. The correlation between stress and anxiety allows our model to

propagate the effect of stressors to fatigue through a more complex but precise

procedure

States of Anxiety:

High

Normal

Sources: [110]

15. Circadian Time State

CEM [103] has proposed several ways to manage the red zone of the circadian

rhythms, which is described as the daily period of lowest energy and alertness,

normally occurring between bedtime and sunrise. Contemporary research has shown

that these periods are often between 3:00 – 5:00 a.m. and p.m. periods during which

the majority of transportation accidents have taken place.

States of Circadian time state:

Drowsy

Awake

Sources: [14] [4]

16. Nutrition

Nutrition is the provision, to cells and organisms, of the materials necessary (in the

form of food) to support life. Many common health problems can be prevented or

alleviated with a healthy diet. In plain words, nutrition can be described as the fuel

that is spent in order to cover an individual‘s bodily needs. Therefore, by adopting a

healthy diet, a seafarer reduces the possibility of poor physical condition and growing

http://en.wikipedia.org/wiki/Cells_%28biology%29

http://en.wikipedia.org/wiki/Organism

http://en.wikipedia.org/wiki/Food

http://en.wikipedia.org/wiki/Life

http://en.wikipedia.org/wiki/Healthy_diet



fatigue. A very common result of a poor diet is anemia, an affection that can cause

various symptoms that impede the individuals overall physical capabilities.

States of nutrition:

Poor

Fair

Sources: [34]

17. Illness

According to medical dictionaries illness is a state of poor health. When the

symptoms and effects of an illness have considerable impact on one‘s organism,

illness is considered a synonym for disease. In our network, we try to demonstrate the

connection between illness and poor diet. Medical research has shown that poor diet

lowers the human organism‘s defenses, making it vulnerable to illnesses. Finally,

illness may come as a result of extreme weather phenomena (such as rain) and

unsuitable clothing for the occasion. The impact of illness on a mariner‘s performance

is severe and it has been recorded that physical capabilities can drop to less than 10%

of the optimum level.

States of Illness:

Yes

No

Sources: various interviews

18. Drug/Alcohol use

For this node we try to represent the impact of drugs (per scripted or non per scripted)

or alcohol use on mariners‘ performance. While there are strict regulations against

such practices, this phenomenon is present on board. These incidents lead in human

errors that are often underreported. The use of alcohol for example can result in errors

http://en.wikipedia.org/wiki/Health

http://en.wikipedia.org/wiki/Disease



in the most simple tasks and procedures. Finally, medications suchas strong

antibiotics can make an individual unable to perform any task given.

States of Drugs/Alcohol use:

Yes

No

Sources: various interviews

19. Physical exercise

Scholars have described physical exercise as any bodily activity that enhances or

maintains physical fitness and overall health or wellness. During the past few years

IMO has proposed the establishment of gymnasiums and other exercise facilities on

board. It is crucial for mariners to maintain a good exercise routine in order to be

ready for the demanding everyday needs while at sea.

States of Physical Exercise:

No

Regular

Sources: [34]

20. Type of task

The Type of Task node describes the nature of the task at hand. During a trip of a

merchant vessel, each member of the crew performs a specific task. At this point, we

divide the tasks in two categories: tedious and normal. Related fatigue surveys have

shown that a monotonous job is more vulnerable to alertness decrease and fatigue.

Such jobs can be the helmsman and watchkeepers in general. The experiencing of

microsleeps during a night shift on the bridge can be found in literature as the

‗helmsman error‘, a result of a tedious procedure and the prevalence of fatigue as

well. The prior probability for this node was also partly based on the questionnaire

responses regarding the nature of their duty and whether they are satisfied with it.

http://en.wikipedia.org/wiki/Physical_fitness

http://en.wikipedia.org/wiki/Health



States of type of task:

Tedious

Normal

Sources: [42] [54] [110]

21. Stress

Almost all research projects mentioned in paragraph 4.4 have indicated that stress is

an important ―party‖ in mariner‘s performance. The variety of stressors that have been

recorded in all these surveys make the quantification of stress an extremely difficult

task to perform. In our model, we list the most important of these factors such as

management pressure and psychological condition.

States of Stress:

High

Normal

Sources: [110] [90]

22. Training



In our network we introduce the training node in order to represent its impact on

stress. Smith [20] has proposed that inadequate training can increase stress levels

significantly, especially in emergency situations. Furthermore, stress can inflict both

sleeping disorders and general physical condition issues.

States of Training:

Poor

Adequate

Sources: [110] [90]

23. Psychological condition

In this node, the effect of an individual‘s general psychological condition on stress is

represented in the CPT. Many interviewees have described that while they experience

an extreme psychological conflict, stress come more easily and their overall

performance is impaired. Many organizations, especially military, perform

psychological test in order to prevent such incidents before the beginning of every

trip.

States of Psychological condition:

Unstable

Stable

Sources: various

6.1.2.2 MANAGEMENT FACTORS

24. Workload

In this node, the workload represents the amount of tasks that an individual has to

process in a given amount of time. The definition of workload is correlated to the time

of rest of a seafarer as well as the opportunities for napping and sufficient sleep. A



quite representative example of heavy workload is the discharging of a VLCC. During

this procedure the bridge officer has to stay awake for almost 40 hours in order to

supervise the discharging. An average time that an adult must stay awake in order to

mitigate fatigue should not exceed 16 hours. Finally, workload also consists the

excessive amount of paperwork that a mariner has to handle while on board.

States of Workload:

Heavy

Normal

Sources: [90] [54]

25. Recent change of shift

In order to abate fatigue of regular night shift operators, shift rotation was introduced

onboard merchant vessels. While this procedure was established for mitigating

fatigue, an abrupt change of day to night shift can seriously disrupt the human

biological clock. In our model, we introduce the Recent change of shift node in order

to show the correlation between this procedure and the effect of the drowsy time

period of the Circadian Rhythms.

States of Recent change of shift:

Yes

No

Sources: [110] [4]



26. High Alert State

The high alert state represents the condition on board when all crew members must

remain stand-by. A typical example could be a fire emergency or when approaching

port. It is clear by the definition of this node that whenever there is such a state the

workload tends to be a heavier than usual and the amount of rest time available

decreases significantly.

States of High Alert State:

Yes

No

Sources: [110]

27. Jet Lag

In the jet lag node we try to show the effect of moving through various time zones in a

short period on the human biological clock. Maritime transport is not as exposed to jet

lag as other transport fields such as aviation, nevertheless we cannot neglect its effect

completely since many mariners have reported related issues. The presence of jet lag

increases the duration of the drowsy time of the circadian clock.

States of Jet Lag:

Yes

No

Sources: [42]

28. Manning Levels

In the past two decades, the prevalence of automation has had a severe impact on

manning levels. While no machine could ever replace a human on-board completely,

ship complements decreased radically. Shipping companies may have benefited from

this new trend but on the other hand workload aboard ships became heavier and



sometimes unbearable. The most common complaint of a seafarer is the fact that

manning levels have a growing tendency to minimize even further. IMO and STCW

have tried to put limitations to this ever-growing trend in order to battle fatigue related

accidents but the fact remains that almost 80% of in-service merchant vessels are

undermanned. In our node we show the connection between manning levels,

workload and management demands.

States of Manning levels:

Low

Normal

Source: [90] [110] [33]

29. Managements Demands

In this node, we summarize the rest of the management and organizational factors

such as company support and pressing schedules. A typical example of a management

related accident could be a pressing deadline which could lead in turn to the violation

of several regulations. A survey [110] has shown that the support of the shipping

company is an efficient measure to battle stress and anxiety.

States of Management Demands:

Demanding

Normal

Sources: [110]



6.1.2.3 SHIP SPECIFIC AND ENVIRONMENTAL FACTORS

30. Work Environment Total

This node represents the overall quality of the work environment on-board. While we

kept the intermediate node from [42] model, we introduced the connection of the

work environment and ship specific elements as the vessel‘s motions and vibrations

which are almost exclusive to the maritime domain. Finally, we make the assumption

that poor quality overall facilities (such as recreational, laundry, galley etc) increase

the probability of fatigue prevalence onboard.

States of Work Environment:

Poor

Fair



Sources: [42] [90] [110]

31. Work Environment Intermediate

The usability of this node is the same as node 8. In order to use the universal transport

node from [42] model we have kept it intact. The parent nodes for this one are light,

noise and temperature capturing the overall nature of a regular work environment.

States of Work Environment Intermediate:

Poor

Fair

Sources: [42]



32. Overall Facilities

In this node we try to capture the overall quality of the vessel‘s facilities. Modern and

ergonomic facilities can provide a pleasant work environment making the mariner‘s

life easier and more pleasant. Factors that may contribute to the overall facilities

quality are the level of automation and the frequency of ship‘s maintenance. It can be

observed in the CPT that we are rather skeptic about the impact of the overall

facilities on the quality of the total work environment.

States of Overall Facilities:

Poor

Fair

Sources: [61]

33. Level of Automation

For this node the following assumption is made:

The higher the level of automation is, the more efficient the overall facilities are.

However, we did not neglect the fact that the more modern and technologically

advance a piece of equipment is, the more complex and difficult is for the average

seafarer to understand its functions, thus the child CPT was filled with cautiousness.

States of Level of Automation:

Poor

Fair

Sources: [14]



34. Age of Vessel

In this node we show that the age of the vessel is a significant factor for the overall

quality of the facilities on-board. The newer the vessel is, the higher the probability

for more ergonomic facilities is. While most vessels that were constructed over 15

years ago are constantly been renovated, several legislations have made mandatory for

new ships facilities such as elevators and better recreational facilities. However, these

installations are rarely part of the renovation of an in-service vessel due to extravagant

costs.

States of Age of Vessel:

2000 minus

2000 plus

Sources: [110]

35. Recreational Facilities

Internet availability, television and several other leisure activities are regarded as the

recreational facilities of a vessel. The importance of these facilities is great since the

maritime work environment is very unique and complex compared to other domains.

Some mariners have to stay months away from home. Therefore, the recreational

facilities help the mariner reduce stress levels and have pleasant time and space to rest

after a tiresome workday. States of Recreational Facilities:

Poor

Normal

Sources: [103]



36. Maintenance

In this node we show the importance of frequent maintenance on the overall quality of

facilities. The lack of maintenance can result in an unhealthy and hazardous working

/living environment aboard ships.

States of Maintenance:

Rare

Frequent

Sources: various

Note: The following nodes are presented synoptically as they are discussed

thoroughly in paragraphs 4.3.1.1. and 4.3.1.2 in order to give emphasis to the

responsibilities of the Naval Architect in fatigue related issues.

37. Sleep Environment

This node describes the overall quality of the vessel‘s environment including

maritime exclusive aspects such as ship‘s motions and vibrations.

States of Sleep Environment:

Poor

Fair

Sources: [23] [14]

38. Sleep Environment intermediate

Taking advantage of the CPT from [42] model, this node functions in a similar way as

node 8 and 30. Contrary to node 37, this one captures the universal nature of sleep

environment neglecting specific aspects discussed above (e.g. vibrations).



States of Sleep environment intermediate:

Poor

Fair

Sources: [42]

39. Vibrations

Vibrations are a common issue onboard. Extreme vibrations levels can result in

hazardous working and sleeping environments and the prevalence of fatigue.

States of vibrations:

Extreme

Normal

Sources: [14] [103]

40. Vessel Motions

Another vessel exclusive fatigue risk factor is the vessel motions. Extreme motions

can make the performing of a specific task extremely difficult and the sleeping

environment unbearable. The main factors that influence the vessel motions are

extreme weather phenomena including waves and tides.

States of Vessel Motions:

Extreme



Normal

Sources: [14] [23]

41. Noise

This node represents the maximum allowed noise levels that a human can endure in

order to avoid several threatening dangers for his general welfare. A noisy

environment can result to fatigue or even more serious issues such as permanent

deafening as well.

States of Noise:

Extreme

Normal

Sources: [14] [96]

42. Temperature

Extreme temperature levels can affect the seafarer‘s performance significantly. There

are specific parts of a ship, such as the engine room, where the high temperatures that

are produced making the environment unhealthy and hazardous for everyone

performing any task under these circumstances.

States of Temperature:

Extreme

Normal



Sources: various

43. Weather

In the weather node we take into consideration every climatic phenomenon that can

affect the living environment onboard. A common example could be the presence of

extreme heaves that influence several hydrostatic and structural aspects of the vessel

thus affecting the ship‘s motions in turn. Our selected states are a correlation of the

Beaufort wind scale and several known sea states.

States of Weather:

Eleven_plus_heavystorm

Ten_eleven_strom

Eight_nine_high_waves

Six_seven_considerable_waves

Four_five_moderate_waves

Three_calm_waves

Calm_no_wave_

Sources: [108]

44. Light

Sufficient lighting is an integral part for a healthy and ergonomic environment. While

sunlight cannot be replaced by any form of artificial lighting, proper illumination on

the workplace can enhance the individual‘s performance and maintain the proper

function of the Circadian rhythms.



States of Light:

Poor

Sufficient

Sources: [14]

45. Humidity

This node captures the importance of humidity in a specific workplace or sleep

environment. Many seafarers have reported this issue as a fatigue risk factor, while

[103] survey states that it is perhaps one of the most important factors regarding an

onboard sleeping environment.

States of Humidity:

High

Normal

Sources: [14] [103]

46. Random Noise

Random noises are those that may serve as a distraction for an individual‘s sleeping

patterns. Announcements made via microphones which address all on-duty crew

members can be an unwanted distraction for those who are sleeping.

States of Random Noise:

Yes

No



Sources: [14] [110]

47. Random Light

Any kind of unwanted light that can distract an individual‘s sleeping continuity. A

typical example is the presence of sunlight during the sleeping time of a night shift

operator at daytime.

States of Random Light

Yes

No

Sources: [14] [110]



CHAPTER 7 BAYESIAN NETWORK

APPLICATION FOR FATIGUE IN

MARITIME SCENARIOS AND

ACCIDENTS



7 BAYESIAN NETWORK APPLICATION FOR FATIGUE IN

MARITIME SCENARIOS AND ACCIDENTS

The scale listed below will serve as an index of our acceptance levels for fatigue

prevalence amongst mariners. Our fundamental assumption is that the seafarers will

eventually suffer the effects of fatigue even if all factors, such as sleeping quality and

work environment, are at a favorable state. Apart from indicating the posterior

probability of our Bayesian Network, the scale introduces the Red Zone ( P(fatigue

=yes > 0.85) where the consequences of fatigue are able to impair the human

performance leading to dangerous outcomes, such as micro-sleeps and serious

decision making errors.

Figure 7. 1: Fatigue Acceptance Scale

In order of provide a more explanative statement of the fatigue scale, we include a

representative examples for each of the scale‘s zones as shown above:

Fresh Zone: Describes the period when an individual feels fresh an active and

his performance is free of the effects of fatigue, (e.g. at the first hours(but not

at the very beginning) of the task at hand in a healthy and ergonomic

environment)

Low Risk: When fatigue is present but rarely affects actions or overall

performance, (e.g. an operator working for a couple of hours with a healthy

background sufficient sleep and physical condition)



Regular Fatigue: The lengthier zone of the whole scale that describes the

period when the individual feels that fatigue starts to be present and slightly

irritating but is regarded as a logical consequence of a day at work (e.g. right

after the 15:00 to 16:00 drowsy period for a clerk working his 9:00 to 17:00

schedule)

Fatigued: At the end of a long day at work when fatigue signs (e.g. yawn or

slow reactions) are obvious and there is a considerable need for rest, ( e.g. at

the end of a day at work where there was no available rest break)

Red Zone: Fatigue has become dominant, reaction times are considerablyslow

and the probability of error is high even for the most simple tasks, (e.g. a sleep

deprived driver making his way home after 20 hours of working in a

hazardous environment while on heavy workload)

Extreme Levels: Describes the period when fatigue has full control over an

individual, making the need for rest a top priority. In this zone the probability

of errors of any kind is almost definite, (e.g. a watchkeeper who falls asleep

while on duty as a result of sleep deprivation, hazardous environment and

exhaustive workload).

7.1 THE IMPACT OF DYNAMIC FACTORS ON FATIGUE PREVALENCE

7.1.1 SUFFICIENT SLEEP AND SLEEP DEPRIVATION

In this paragraph we are going to demonstrate the impact of sleep deprivation on

fatigue related issues. Throughout the entire body of our research, we realized that the

overall sleeping quality of an individual is the most important contributing factor to

the emergence of fatigue. Furthermore, scholars have indicated that the critical point

among sufficient sleep and deprivation, for an average adult, is the period between six

to eight hours of uninterrupted sleep. While this observation might not apply to every

single adult, since other factors such as age and physical condition take part, it has

been accepted as a fundamental assumption when studying fatigue.

In order to show this particular interaction between fatigue and adequate sleep, we

modeled our Bayesian Network in such manner that allows us to import various

sleeping quantity inputs. While most nodes contain only two states, the hours awake

node provides seven states, with their conditional probabilitie,s for the user to import

as evidence. It is important to mention at this point that prior and conditional

probability tables indicate that most mariners sleep between six to seven hours in a

normal workday, while these hours are significantly reduced in a heavy workload day

with the average values accumulated around six hours.

The following table and figure contains the results of the posterior probability for the

fatigue node for each of the Hours Awake states that are imported as evidence. Each

time we import evidence, we update the network in order to set the evidence node

state as 1 and then propagate the information for the calculus of the rest of the

posterior probabilities. Note: From now on, evidence will be presented as (e).



Table 7. 1: Fatigue - Hours Awake

11.Hour Awake State (e) 1.Fatigue Posterior Probability P( Fatigue = yes) % Change

Twelve (12) 0,635 0,00

Fourteen (14) 0,643 1,26

Sixteen (16) 0,676 5,13

Eighteen (18) 0,783 15,83

Twenty (20) 0,838 7,02

Twenty-four (24) 0,850 1,43

More than a day (24+) 0,858 0,94

Note: When More than a day is imported, the sleep quantity node intakes immediately

the deprived state almost in a similar manner as we import evidence. In plain words,

this means that whenever an individual has not slept for more than twenty four hours

it is definite that he/she will be sleep deprived.

Figure 7. 2: Fatigue - Hours Awake

0,6

0,65

0,7

0,75

0,8

0,85

0,9

0,95

1

10 12 14 16 18 20 22 24 26 28

P(F

ati

gu

e=yes

)

Hour awake (h)

Fatigue - Hours awake



The figure above indicates that the critical limit of prevalence, which is set as P

(Fatigue= yes) = 0.85, is surpassed when Hours Awake exceed 20. The abrupt turn

from 16 to 20 pinpoints the fact that the thin line between enough sleep and sleep

deprivation rests between 6-8 hours of uninterrupted sleep. When the workload

aboard is heavy, there is rarely the chance for more than 6 hours of sleep.

Furthermore, accidents and investigations reports do not seem to identify the

significance of napping, which scientists describe as crucial in workplaces where

sleeping patterns are easily disturbed. If a mariner manages to sleep more than eight

hours a day, even for a small number of consecutive days, then fatigue probability

decreases significantly. The same occurs for sleep deprivation as well.

7.1.2 WEATHER AND SEA STATE

The maritime environment is a unique workplace in comparison to any other transport

domain. In our attempt to provide a representative modeling tool for such an

inimitable environment, we created a network that considers weather and sea state as

an integral part of the problem.

The science of Naval Architecture and Marine Engineering has always focused on the

interaction of a vessel‘s hydrostatics, hydrodynamics, overall structural endurance

withthe variety of weather phenomena. Though various sources of literature and

interviews with experts and mariners, it was found that the sea state affects directly

both the working and sleeping environment, thus making it a common fatigue related

factor.

In order to demonstrate this interaction we have introduced the weather node in

similar manner to the Hours Awake one. Using knowledge from both scientific

expertise and statistical data, we applied our experience and judgment into creating a

node that connects weather and the vessel‘s motions. While some probabilities are

rough estimations provided by experts, this node succeeds in representing the

connection between the ship‘s dynamic movement and a bad weather scenario.

Making the assumption that the vessel‘s motions are primarily affected by waves that

create specific sea states environments‘, we used the Beaufort scale in order to

correlate these two parties, as listed below.

Note: Both literature and experts agree that States 1) and 2) are rather rare, a fact

reflected in their prior probabilities that are significantly low. 12+ values in the

Beaufort scale refer to phenomena such as hurricanes and typhoons.

Figure 7.3 is the presentation of the Beaufort scale. On the third row we can see the

mariner‘s term for each stage of the scale while on the fifth indicates a list of effects

at land relating to wind speed.



Figure 7. 3: The Beaufort scale

Our weather node has seven states as presented below:

1) Eleven Plus – Heavy storm (hurricane)

2) Ten to Eleven – Severe Storm

3) Eight to Nine – High Waves

4) Six to Seven – Considerable Waves

5) Four to Five – Moderate Waves

6) Two to Three – Calm Waves

7) One to Two – Calm Sea

In the following table and figure, we present the results of the posterior probability, P

(Fatigue = Yes), for each Weather node state that was imported as evidence.

Note: The conditional probability table of the Vessel Motions node was arranged

taking into consideration the impact of weather on the movements of a merchant ship

of average size such as a Bulk Carrier of 10000-40000 tones DWT (deadweight).



Table 7. 2 : Fatigue -Weather

43.Weather Node states(e) 1.Fatigue Posterior Probability P( Fatigue = yes) % Ch.5

Eleven Plus – Heavy storm 0,84 0,00

Ten to Eleven – Severe stor. 0,837 0,36

Eight to Nine – High Waves 0,824 1,58

Six to Seven – Considerable

Waves 0,801 2,87

Four to Five – Moderate

Waves 0,78 2,69

Two to Three – Calm Waves 0,773 0,91

One to Two – Calm Sea 0,763 1,31

Figure 7. 4: Fatigue - Weather

In the figure above the passage from moderate calm sea to considerable waves (three

to seven) is represented by a rapid abrupt increase on the fatigue probability. The

posterior probability not reaching the red zone is due to the fact that at this point the

weather node serves as the only evidence of our rather ‗Good will‘ network. It is

crucial to mention that early in this chapter we realize that sleep is proven to be more

influential than any other contributing factor.

5 % change of P( Fatigue =Yes)

0,74

0,76

0,78

0,8

0,82

0,84

0,86

0 2 4 6 8 10 12 14

P (

Fa

tig

ue=

yes

)

Wind Velocity (B)

Fatigue -Weather



7.2 MARITIME ACCIDENTS

7.2.1 EXXON VALDEZ

The Exxon Valdez oil spill occurred in Prince William Sound, Alaska, on March 24,

1989, when the Exxon Valdez, an oil tanker bound for Long Beach, California, struck

Prince William Sound's Bligh Reef and spilled 260,000 to 750,000 barrels (41,000 to

119,000 m3) of crude oil. It is considered to be one of the most devastating human-

caused environmental disasters ever to occur in history. As significant as the Valdez

spill was — the largest ever in U.S. waters until the 2010 Deepwater Horizon oil spill

— it ranks well down on the list of the world's largest oil spills in terms of volume

released. However, Prince William Sound's remote location, accessible only by

helicopter, plane and boat, made government and industry response efforts difficult

and severely taxed existing plans for response.

Figure 7. 5: The Exxon Valdez oil spill

Apart from identified as one of the most disastrous accidents of all time, the Exxon

Valdez oil spill is also the most common example of Human Factor related incidents

in the maritime domain. Even if the spill took place over twenty years ago, when

double-hulls were not yet mandatory for oil tankers, scientists still investigate the

incident in order to provide safer vessels and procedures and eventually prevent

similar catastrophes in the future [66].

http://en.wikipedia.org/wiki/Prince_William_Sound

http://en.wikipedia.org/wiki/Exxon_Valdez

http://en.wikipedia.org/wiki/Oil_tanker

http://en.wikipedia.org/wiki/Long_Beach,_California

http://en.wikipedia.org/wiki/Prince_William_Sound

http://en.wikipedia.org/wiki/Bligh_Reef

http://en.wikipedia.org/wiki/Oil_spill

http://en.wikipedia.org/wiki/Crude_oil

http://en.wikipedia.org/wiki/Environmental_disaster

http://en.wikipedia.org/wiki/Deepwater_Horizon

http://en.wikipedia.org/wiki/Deepwater_Horizon_oil_spill

http://en.wikipedia.org/wiki/Oil_spills#Largest_oil_spills



In this paragraph we will try to apply our fatigue-Bayesian Network on the Exxon

Valdez incident through the import of evidence that was provided from the accident‘s

report. Furthermore, we will provide an alternate step-by-step scenario of what the

fatigue posterior probability would be if all the evidence had the preferred state or the

states proposed by the legislation party (e.g. STCW).

Note1: At this point, we will investigate the 3rd

mate‘s (Cousins) actions and

background, who was the bridge watchkeeper at the time of the grounding.

Note 2: The vessel motions conditional probability table was slightly altered because

of the ship‘s size. The changes that were made indicate the fact that weather

phenomena have lesser impact on a 200000 DWT tanker‘s motions than it would have

on 10000 DWT Bulk Carrier or General Cargo ship.

The following findings are part of the Final Report, submitted by the Alaska Oil Spill

Commission, published in February 1990 by the State of Alaska [X]:

7.2.1.1 ACTUAL INCIDENT

1. Manning Levels: Low (e1)

„Cousins was the only officer on the bridge-a situation that violated company policy

and perhaps contributed to the accident. A second officer on the bridge might have

been more alert to the danger in the ship's position, the failure of its efforts to turn,

the autopilot steering status, and the threat of ice in the tanker lane‟

The 3rd

Mate, Cousins, was left alone on the bridge against company‘s policies and

legislation. It is important to mention that if the 3rd

mate had better knowledge

regarding fatigue, he would probably identify the risk of staying on the bridge alone.

2. Type of Task: Tedious (e2)

Knowing the nature of the job of a watchkeeper, we assumed the tedious nature of the

task at hand.

3. Management Demands: Demanding (e3)

„Minimum vessel manning limits are set by the U.S. Coast Guard, but without any

agency wide standard for policy. The Coast Guard has certified Exxon tankers for a

minimum of 15 persons (14 if the radio officer is not required). Frank Iarossi,

president of Exxon Shipping Company, has stated that his company's policy is to

reduce its standard crew complement to 16 on fully automated, diesel-powered

vessels by 1990. "While Exxon has defended their actions as an economic decision",

the manning report says, "criticism has been leveled against them for manipulating

overtime records to better justify reduced manning level”.

In addition to the 3rd

Mate left alone on the bridge, it seems that Exxon Valdez was

undermanned for a vessel of its type and size.



4. Hours Awake: Twenty (e4)

―By this time Third Mate Cousins had been on duty for six hours and was scheduled

to be relieved by Second Mate Lloyd LeCain. But Cousins, knowing LeCain had

worked long hours during loading operations during the day, had told the second

mate he could take his time in relieving him. Cousins did not call LeCain to awaken

him for the midnight-to-4-a.m. watch, instead remaining on duty himself.”

“Cousins' duty hours and rest periods became an issue in subsequent investigations.

Exxon Shipping Company has said the third mate slept between I a.m. and 7:20 a.m.

the morning of March 23 and again between 1: 30 p.m. and 5 p.m., for a total of

nearly 10 hours sleep in the 24 hours preceding the accident. But testimony before the

NTSB suggests that Cousins "pounded the deck" that afternoon, that he did

paperwork in his cabin, and that he ate dinner starting at 4:30 p.m. before relieving

the chief mate at 5 p.m. An NTSB report shows that Cousins' customary in-port

watches were scheduled from 5:50 a.m. to 11:50 a.m. and again from 5:50 p.m. to

11:50 p.m. Testimony before the NTSB suggests that Cousins may have been awake

and generally at work for up to 18-20 hours preceding the accident.‖

It is important to mention that if the 3rd

mate had had better knowledge regarding

fatigue, he would probably have identified the risk of staying awake for that amount

of time. Instead he chose to relieve the 2nd

mate, who would have probably been

affected by fatigue as well, making himself vulnerable to errors and

underperformance.

From the report‘s excerpt listed above we can assume that the 3rd

Mate was awake for

almost twenty hours prior the incident.

Table 7. 3: Exxon Valdez Actual

Evidence 1.Fatigue Posterior Probability P( Fatigue = yes)

no evidence 0,776

e1 0,81

e2 0,855

e3 0,859

e4 0,883

At this point we feel obliged to mention that the 3rd

mate was not supposed to stay

alone in the bridge. The captain, who the investigations showed that he was under the

influence of alcohol, was sleeping at the time of the incident and even when the



situation became out of hand, he returned to the bridge but made a series of mistakes

that led to the magnitude of this disaster.

The posterior probability change from (e2) to (e3) is relatively low due to the fact

that:

The prior probability distribution of the Management Demand node is in favor

of the Demanding state since schedules are often pressing aboard merchant

ships.

The red zone has already been reached by the Type of Task evidence.

We must also mention that the type of task seems to be of great importance even in

this early point of our network testing.

Figure 7. 6: Exxon Valdez Actual Incident

7.2.1.2 OPTIMAL ALTERNATE SCENARIO

1) Manning Levels: Normal (e1)

2) Type of work: Normal (e2)

We decided to change this node‘s state in order to present the possibilities of the

watchkeeper‘s job to become more interesting and less monotonous.

3) Management Demands: Normal (e3)

4) Hours Awake (e4)

Eight Hours of uninterrupted sleep suggested by the STCW 75 Sixteen Hours

awake within a day.

0,72

0,74

0,76

0,78

0,8

0,82

0,84

0,86

0,88

0,9

no evidence e1 e2 e3 e4

P (

Fa

tig

ue

=y

es)

Given evidence

Exxon Valdez Actual incident



Table 7. 4: Exxon Valdez Alternate scenario


no evidence 0,776

e1 0,717

e2 0,695

e3 0,663

e4 0,612

The table above indicates that the best way for the posterior probability to drop is

through sufficient sleep (e4) and less monotonous activities.

Figure 7. 7: Exxon Valdez Alternate scenario

0,6

0,62

0,64

0,66

0,68

0,7

0,72

0,74

0,76

0,78

0,8

no evidence e1 e2 e3 e4

P (

Fa

tig

ue

=y

es)

Evidence given

Exxon Valdez Alternate scenario



Figure 7. 8: Exxon Valdez scenario Comparison

The figure above indicates the following fact:

By providing a better and healthy framework, the posterior probability would

decrease more than it would increase if the environment was equally

hazardous.

7.2.2 MFV BETTY JAMES

In this paragraph we will be investigating the grounding of Fishing Vessel Betty

James (130 G.T.) on the evening of 9 July 2000. Our main resource of information

was the Report No 34/2000 conducted by MAIB, (MAIB). The following excerpt was

taken from the synopsis of this investigation [56].

“Mfv Betty James landed her catch in Mallaig, Scotland, on the evening of 9 July

2000, then sailed at 0015 the following day to return to the fishing grounds. At 0230

she ran aground. The person on watch had fallen asleep and a planned alteration of

course to take the vessel between the isles of Rhum and Eigg was missed. A watch

alarm was fitted and working, but it failed to wake either the watchkeeper or the crew

asleep below in the accommodation.”

0,6

0,65

0,7

0,75

0,8

0,85

0,9

no

evidence

e1 e2 e3 e4

P (

Fa

tig

uE

= y

es)

Exxon Valdez Comparison

Actual

Alternate



Figure 7. 9: The grounding of Betty James

At this point, we will investigate the actions and performance of the watchkeeper who

fell asleep while on duty on the bridge. After studying thoroughly the report provided

by MAIB we were able to identify the follow findings that will serve as evidence for

our network. Apart from importing evidence of the actual incident we will also

provide an alternate Good Will scenario similar to that of paragraph 7.2.1.


1) Drug? Alcohol Use: Yes (e1)

―The watchkeeper had consumed three bottles of beer while in Mallaig. Three bottles

of beer is a moderate quantity of alcohol to consume, which under normal

circumstances might have had only a negligible effect on Peter Matheson. However,

given that in the previous seven days he had not consumed any alcohol nor had any

periods of sleep greater than 4 hours, duration, it is possible that the alcohol

consumed by Peter Matheson had a greater soporific effect than he realized‖.

Following the investigation findings we can assume that alcohol consumption had an

impact on the operator‘s performance. It is important to highlight the fact that the

watchkeeper overestimated his endurance to the effects of alcohol, a typical practice

amongst mariners or individuals in general.


“Wheelhouse practices and ergonomics allowed him to conduct his watch while

seated, and kept him inactive”



Apart from the already monotonous nature of watchkeeping, the operator remained

seated and inactive, (reading newspapers).

3. Manning Levels: Low (e3)

„He was alone in the wheelhouse‟

4. Work Environment: Poor (e4)

“The presence of a television, video recorder and domestic radio in the wheelhouse

encouraged a recreational, rather than working, environment”

While our network takes into consideration different factors in order to define the

quality of a workplace, we can make the assumption that a distracting environment is

a poor one.

5. Hours Awake: 18 Hours (e5)

„He probably had no more than 6 hours of sleep in the previous 24.‟

6. Sleep Disorders: Regular (e6)

„He had experienced a disrupted sleeping pattern since the vessel sailed on 3 July‟

Since the grounding occurred on 9 July, we can assume that the watchkeeper had

sleep disorder for almost a week prior the incident.

Note 1: We have altered the Vessels motions CPT in similar manner to the Exxon

Valdez example because of the ship‘s size (L = 16.49 m). The changes were minor

because even though bad weather may had a more significant impact on a small

vessel, ships of that type and size do not sail in open waters where extreme weather

phenomena might appear.

Note 2: Although the actual incident took place at 2:30, a period very close to the 3:00

to 5:00 drowsy time of the circadian rhythms, we decided not to import evidence on

this node, as the watchkeeper might have been asleep for some time.

Once again, the monotonous nature of watchkeeping is enough to drive the posterior

probability to the red zone almost solely by itself. Furthermore, it is obvious that even

if music and leisure activities can be an effective way to ameliorate the mariners‘

morale, in this case it proved to be a distracting factor producing a hazardous work

environment instead of a countermeasure for tediousness.



Table 7. 5: Betty James Actual incident


no evidence 0,779

e1 0,801

e2 0,854

e3 0,87

e4 0,924

e5 0,926

e6 0,936

Figure 7. 10: Betty James Actual incident

In this incident, the 0.85 boundary/red zone is surpassed rapidly, early in the evidence

importing process. Another interesting finding is that even if the watchkeeper had not

had sufficient sleeping time prior to the accident, the impact of sleep deprivation was

minimal since extreme fatigue levels had already been reached.

0,7

0,75

0,8

0,85

0,9

0,95

no evidence e1 e2 e3 e4 e5 e6

P (

Fa

tig

ue

=y

es)

Evidence given

Betty James actual




1) Drug/Alcohol Use: No (e1)

2. Type of Task: Normal (e2)

In similar manner to that of Exxon Valdez.

3. Manning Levels: Normal (e3)

4. Work Environment: Fair (e4)

At this point we propose a non distracting environment importing soft evidence of

P (W.E. = Fair) = 0.8.

5. Hours Awake: 16 Hours (e5)

Proposed by STCW 75.

6. Sleep Disorders: Rare (e6)

Table 7. 6: Betty James Alternate scenario


no evidence 0,779

e1 0,778

e2 0,763

e3 0,698

e4 0,631

e5 0,55

e6 0,535

The table above proves a basic truth about watchkeeping. If there was an additional

watchkeeper on duty (mandatory for almost every flag state these days) the accident

may had never happened, a fact easily made out by the transition of (e2) to (e3).

Finally, sufficient sleep, as proposed by STCW, drops the probability almost 8% and

into the fresh zone. The 30% drop of the posterior probability again proves that it is

easier to make the situation better in the similar way it would get worse.



Figure 7. 11: Betty James alternate scenario

Figure 7. 12: Betty James scenario comparison

0,4

0,5

0,6

0,7

0,8

0,9

1

no

evidence

e1 e2 e3 e4 e5 e6

P (

Fa

tig

ue

=y

es)

Evidence given

Betty James alternate scenario

0,4

0,5

0,6

0,7

0,8

0,9

1

no

evidence

e1 e2 e3 e4 e5 e6

P (

Fa

tig

ue=

yes

)

Betty James comparison

Betty James actual

Betty James alternate



7.2.3 ANTARI (GENERAL CARGO 2446 G.T.)

In a similar approach to the preceding paragraph, at this point we are investigating the

grounding of General Cargo vessel Antari (2446 G.T.). Once again, our main resource

of information was the Report No 7/2009 February 2009 conducted by MAIB. The

following excerpt was taken from the synopsis of this investigation [55]:

“At 0321 on 29 June 2008 the general cargo vessel Antari grounded on the coast of

Northern Ireland, while on passage from Corpach, Scotland to Ghent, Belgium. The

officer of the watch had fallen asleep shortly after taking over the watch at midnight

when the vessel was passing the peninsula of Kintyre (Scotland). With no-one awake

on the bridge, the vessel continued on for over 3 hours, crossing the North Channel of

the Irish Sea before grounding on a gently sloping beach about 7 miles north of

Larne”.

Figure 7. 13: The Antari Grounding

The analysis of the findings and the import of evidence in our network will be part of

the same procedure followed in the preceding examples.

Note 1: The Vessel Motion CPT was altered in similar manner to the preceding

example, the 6000 DWT General Cargo has greater endurance to waves than the

Fishing Vessel discussed earlier while it is also mentioned in the report that its main

voyages are within coastal waters due to the nature of short-trade shipping routing.


1. Random Noise: Yes (e1)

“In the 24 hours preceding the grounding, his sleep pattern had been disturbed by

the port call and cargo operations…”.

It is implied that the watchkeeper‘s sleep was disrupted by random noises.



2. Circadian Rhythms Intermediate: Drowsy (e2)

“….although he had the opportunity to rest, his circadian rhythm was probably

sufficiently disrupted to prevent him obtaining adequate rest during this period…”.

Our interpretation of the fact mentioned above is that the watchkeeper was in the

drowsy period of his Biological Clock. While, the report indicates that he might have

experienced difficulties to obtain sufficient sleep, our network has the ability to take

advantage of the evidence regarding Circadian Rhythms separately.

3. High Alert State: Yes (e3)

“Antari was trading around the North West coast of Europe and had made 21 port

calls in the 2 months preceding the accident. Every port call required the master‟s

and chief officer‟s intense involvement: preparations for arrival and departure,

pilotage, supervision of cargo operations, official and cargo paperwork. Audits and

statutory inspections are normally undertaken in port, and these additional demands

cannot normally be contained within the 6 hours on/6 hours off watch pattern since

they frequently occur during what would otherwise be regarded as rest periods”.

The frequency of port calls resulting in exhaustive workload as well as limited time

for rest and sleep has led into importing evidence in the High Alert State node. This

time we have no specific information about hours of sleep or manning levels

(although the watchkeeper was alone in the Bridge, that was not against the legislation

of the flag state) but the evidence set in this state updates these nodes by taking

advantage of the capabilities that a Bayesian network provides each user ( High alert

state heavy workload hours awake).


“The atmosphere on the bridge would have been very quiet, and once he sat in the

chair, with no lookout posted and the watch alarm turned off, there were no means of

anyone knowing that the chief officer had fallen asleep”.

Apart from the already monotonous nature of watchkeeping, the operator again

remained seated and inactive.

5. Recent Change of Shift: Yes (e5)

“The master‟s records show that he worked the 6 to 12 watch regardless of whether

or not he was actually working. In the case of the chief officer, the record shows that

he was working between midnight and 0600 on 28 June, when it is known that he was



sleeping from 0200 until about 0700. Thus, they do not accurately reflect the change

of routine experienced in port”.

It can be implied that the watchkeeper had experienced a recent change of shift due to

low manning levels (accepted by the flag state) and the nature of short sea trading.

Table 7. 7: Antari actual incident


no evidence 0,777

e1 0,78

e2 0,86

e3 0,888

e4 0,913

e5 0,918

Figure 7. 14: Antari actual incident

The main finding of the figure above is the recognition of the Circadian Rhythms

disruption as one of the most dangerous factors that lead to fatigue prevalence. The

fact that the posterior probability does not seem to be affected by the import of (e5) is

the result of the already disrupted biological clock. If there was no evidence regarding

Circadian Rhythms the impact of abrupt shift rotation would be greater than shown

0,75

0,77

0,79

0,81

0,83

0,85

0,87

0,89

0,91

0,93

no evidence e1 e2 e3 e4 e5

P (

Fa

tig

ue

=y

es)

Evidence given

Antari actual



above. For example if (e5) was the only evidence available, the updating of the

network would produce a posterior probability of 0.848, which would be a significant

change compared to the 0.777 prior probability.


1. Random Noise: No (e1)

Through better noise dampening and isolation techniques.

2. Circadian Rhythms Intermediate: Awake (e2)

Through Higher manning levels and better shift rotation.

3. High Alert State: No (e3)

Through improved route planning and management policies.

4. Type of Task: Normal (e4)

Similar to 7.2.1. And 7.2.2. A kind of practice that would keep the watchkeeper

active and awake (e.g. music) while on duty, especially at night shifts.

5. Recent Change of Shift: No (e5)

Through similar practices proposed for (e2) and (e3)

Table 7. 8: Antari alternate scenario


no evidence 0,777

e1 0,776

e2 0,764

e3 0,762

e4 0,746

e5 0,73

The small change of the posterior probability after the import of all five evidences

reflects that their prior distribution (for the fatigue mitigation states) was already low.



Figure 7. 15: Antari alternate scenario

In comparison to the actual incident scenario, the above graph indicates that even by

making sure that all these factors are controlled, the posterior probability does not

change dramatically. On the other hand, when these factors are at their hazardous

states the fatigue posterior probability surpasses acceptable levels rather quickly while

reaching extreme values close to 0.92. This observation is also visible in the following

figure where the Actual Incident and the Alternate Scenario are presented together.

Figure 7. 16: Antari scenario comparison

7.3 HYPOTHETICAL SCENARIO OF 30000 G.T. CONTAINERSHIP

0,7

0,71

0,72

0,73

0,74

0,75

0,76

0,77

0,78

0,79


P (

Fa

tig

ue

=y

es)

Evidence given

Antari alternate scenario

0,7

0,75

0,8

0,85

0,9

0,95

no

evidence

e1 e2 e3 e4 e5

P (

Fa

tig

ue

= y

es)

Antari comparison

Antari actual

Antari alternate



In this paragraph we discuss a hypothetical scenario of a 30000 G.T. containership in

order to demonstrate our model‘s full potential and capabilities. In the previous

paragraphs factors such as sleep deprivation and shift rotation were mentioned

numerous times. Unfortunately, factors regarding physical condition and working

environment are rarely encountered in accident reports.

Container ships are considered high-risk vessels due to the nature of their cargo and

work schedules. Some of the risks are linked to the loading and unloading of

containers. The risks involved in these operations affect both the cargo being moved

onto or off the ship, as well as the ship itself. Containers, due to their fairly

nondescript nature and the small number handled in major ports, require complex

organization to ensure they are not lost, stolen or misrouted. In addition, as the

containers and the cargo they contain make up the vast majority of the total weight of

a cargo ship, the loading and unloading is a delicate balancing act, as it directly

affects the centre of mass for the whole ship.

In our hypothetical scenario we try to interpret the complex nature of a typical

containership into a representative fatigue inducing example. Some common facts

regarding containerships are listed below:

High service speed

Low Wave Endurance

Low Block Coefficient

Low Friction Resistance

Regular Port Calls and Heavy Workloads

Figure 7. 17: An unfortunate incident aboard a containership

http://en.wikipedia.org/wiki/Cargo



7.3.1 PROPOSED EVIDENCE

In similar approach to the previous examples, we will try to present a possible

scenario regarding a watchkeeper working his shift in the bridge. The proposed

evidence for our hypothetical example are listed and explained below:

1) Overweight watchkeeper Physical exercise: No (e1)

Nutrition: Poor (e2)

A study conducted on Danish seafarers [34] provided the following results:

“The study comprised 1257 male seafarers. There were statistically significant more

overweight seafarers in all age groups compared to a reference group ashore. Among

those between 45 and 66 years of age 0.7% had a weight below normal, 22.7% had

normal weight and 76.6% had a weight above normal, while 30.9% of this age group

was obese”.

These findings show that overweight issues are very common amongst marines so our

proposing of this evidence is realistic enough.

2) Aged Watchkeeper Age: 45 plus (e3)

For this node we propose a watchkeeper who is over forty years of age. Given the fact

that several years of service are required in order to reach the rank of Captain, First,

Second or even Third Mate, a watchkeeper can easily be over forty five years old, a

fact also confirmed by the [110] questionnaire.

3) Lack of Quality Recreational Facilities Recreational Facilities: Poor (e4)

According to ABS Crew Habitability Guide [61], mariners consider the following as

vital parts of an ambient recreational environment:

– Internet access

– Access to a telephone

– Able to send/receive email

– Exercise equipment

– Satellite Television

– Library (DVDs & books)

The same guide also reports that better onboard living conditions (habitability and

recreation facilities) as one of the main concerns when living at sea for long periods.

In conclusion, we can assume that since this matter is still a common request amongst

mariners, there is substantial probability that the overall quality of recreational

facilities onboard is poor. Internet availability was recently made mandatory for



Greek merchant vessels and thought to be a great step forward. In reality, the

importance of communicating with his social circle in the land is crucial for a mariner

and his morale concluding that shipping companies do not yet approach the issue

correctly.

4) High Humidity levels Humidity: High (e5)

Crew Endurance Management system [103] identifies high humidity as a common

issue aboard merchants vessels while it presents it as the main cause for Heat

Exhaustion. The following excerpt is part of the chapter regarding sleep and circadian

rhythms:

“Energy is restored optimally during uninterrupted sleep periods of 7 to 8 hours, with

the sleeper reclined on a comfortable mattress in dark and quiet room at 65 - 70° F

and 60 - 70% humidity”.

5) Exhaustive workload, regular port calls and sleep deprivation

Management Demands: Demanding (e6)

Hours Awake: Twenty four (e7)

The following passages by Martin Machado (AB seaman)6 are part of his testimony

regarding his months of service on a container ship:

“I have been living and working as a deckhand on a 906 foot container ship making

57 day runs from New York to Singapore, while hitting many ports in between”.

―Overtime is where a sailor makes his money, so we take as much as they'll give. I

typically get around 12 hours work each day at sea, and in port I can work almost 24

hours straight at times. So any sleep is much appreciated”.

Note: At this point we could also import evidence regarding heavy workload and high

alert state but decided against it since the Hours Awake evidence affect these nodes

directly thus producing a realistic scenario nevertheless.

6) Bad weather Weather: six_seven_considerable_waves (e8)

The Containershipping.nl website‘s containerships accident database indicates that in

half of the 42 incidents that were investigated the last decade, weather was a major

factor. A typical example is that of the containership MSC Napoli:

―While en route from Belgium to Portugal, on January 18, 2007, during European

windstorm Kyrill, severe gale force winds and huge waves caused serious damage to

6 Able Seamen‘s duties include standing watch as helmsman and lookout

http://en.wikipedia.org/wiki/Belgium

http://en.wikipedia.org/wiki/Portugal

http://en.wikipedia.org/wiki/European_windstorm

http://en.wikipedia.org/wiki/European_windstorm

http://en.wikipedia.org/wiki/Kyrill_%28storm%29

http://en.wikipedia.org/wiki/Watchstanding

http://en.wikipedia.org/wiki/Helmsman

http://en.wikipedia.org/wiki/Lookout



Napoli's hull, including a crack in one side and a flooded engine room. The ship was

then 50 miles (80 km) off the coast of The Lizard, Cornwall”.

While we avoided importing evidence that would reflect an extreme weather incident,

we felt that the six to seven B (Beaufort scale) sea state would be a representative

example of a ‗bad‘ weather scenario.

7.3.2 CALCULATING THE PROBABILITY FOR THE HYPOTHETICAL

SCENARIO

7.3.2.1 ACCIDENT FRIENDLY SCENARIO

Table 7. 9: Containership accident friendly scenario

Evidence 1.Fatigue Posterior Probability P( Fatigue = yes) Change

(%)

No evidence 0,777 0.00

e1 0,789 1.54

e2 0,794 0.63

e3 0,805 1.39

e4 0,845 4.97

e5 0,859 1.66

e6 0,878 2.21

e7 0,91 3.64

e8 0,913 0.33

Even though the first three evidence do not seem to affect the posterior probability

significantly, (e4) reaches the red zone by an almost 5% increase. This fact can be

easily explained as the vessels recreational facilities is a micrographic representation

of the vessel‘s overall facilities (as shown by the parent-child connection in our

network). Another interesting finding is that weather seems to leave the posterior

probability unnaffected due to the fact that (e8) was imported last when the situation

was already extreme and could hardly worsen.

http://en.wikipedia.org/wiki/The_Lizard

http://en.wikipedia.org/wiki/Cornwall



Figure 7. 18: Containership accident friendly scenario

7.3.3.2 OPTIMAL SCENARIO

1) Fit watchkeeper Physical exercise: Regular (e1)

Nutrition: Fair (e2)

Through regular exercise and proper regulation regarding the mariners‘ physical

condition.

2) Young watchkeeper Age: 45 minus (e3)

3) High Quality Recreational Facilities Recreational Facilities: Fair (e4)

By embracing the mariners‘ needs and complaints.

4) Low Humidity levels Humidity: Low (e5)

5) Normal workload, improved shift planning, adequate sleep

Management Demands: Normal (e6)

Hours Awake: Sixteen (e7)

6) Normal weather Weather: three_calm__waves (e8)

Again, in similar manner to the Antari alternate scenario, after importing the first five

evidences the posterior probability stays relatively high, due to the low prior

distribution of the nodes‘ positive states. Finally, once again, sufficient sleep and

0,76

0,78

0,8

0,82

0,84

0,86

0,88

0,9

0,92

0,94

no evid. e1 e2 e3 e4 e5 e6 e7 e8

P (

Fa

tig

ue

=y

es)

Evidene given

Containership Accident Friendly Hypothetical

Scenario



proper management, manning and shift rotation seem to provide a quite improved

context for battling fatigue. Weather on the other hand does not seem to be a major

concern in such a healthy and ergonomic workplace.

Table 7. 10: Containership optimal scenario

Evidence 1.Fatigue Posterior Probability P( Fatigue = yes) Change (%)

No evidence 0,777 0.00

e1 0,76 2.24

e2 0,758 0.26

e3 0,753 0.66

e4 0,725 3.86

e5 0,719 0.83

e6 0,656 9.60

e7 0,566 15.90

e8 0,559 1.25

Figure 7. 19: Containership optimal scenario

0,5

0,55

0,6

0,65

0,7

0,75

no evid. e1 e2 e3 e4 e5 e6 e7 e8

P(F

ati

gu

e =

yes

)

Evidence Given

Containership Optimal Hypothetical scenario



Figure 7. 20: Containership Scenario Comparison

0,5

0,55

0,6

0,65

0,7

0,75

0,8

0,85

0,9

0,95

1

no

evid.

e1 e2 e3 e4 e5 e6 e7 e8

P (

Fa

tig

ue

=y

es)

Containership Scenario Comparison

Accident friendly

Best case



CHAPTER 8 SENSITIVITY ANALYSIS



8 SENSITIVITY ANALYSIS

8.1 INTRODUCTION

Sensitivity analysis can be described as a method that defines how ―sensitive‖ a

mathematic model really is when the parameters and the overall structure of the

model experience changes. The primary advantage that a sensitivity analysis provides

is the studying of the uncertainties of parameters that are almost impossible to

measure accurately in the real world.

This procedure helps to build confidence in the model by solving such precision and

quantification issues thus improving the model‘s reliability. Therefore, when building

a system dynamics model, the modeler is usually at least somewhat uncertain about

the parameter values he chooses and must use estimates or expert knowledge.

Sensitivity analysis allows the researcher to find the accuracy needed for an accurate

and functional model. If the analysis shows signs of insensitivity to parameter change,

then it is probable that a rough estimation might serve as better information than an

actual value. Sensitivity analysis can also pinpoint which parameter values are

reasonable to use in the model. The ultimate objective of this procedure is to create a

model that reflects the real world so the best way to determine that is by comparing

the result to realistic scenarios. If these comparisons show similar findings then the

model has both high precision and functionality.

Finally, sensitivity analysis can also provide knowledge about the overall tests

dynamics of a system. Experimenting with a wide range of values can offer insights

into behavior of a system in extreme situations. Through these experimentations the

modeler can track down where the system behavior greatly changes after a specific

parameter alters. To conclude, this method identifies a leverage point in the model, a

parameter whose specific value can significantly influence the behavior mode of the

system [10].

8.2. BAYESIAN NETWORK SENSITIVITY ANALYSIS

8.2.1 PARAMETER DESICION

In this paragraph we perform a sensitivity analysis on our Bayesian Network

regarding fatigue in the maritime domain. Through this procedure, we hope to test our

model in reliability, precision and ‗real world‘ reflection (see paragraph 8.1). In the

previous chapter our network was used in order to calculate the posterior probability

of fatigue in real accidents and hypothetical scenarios with the use of evidence. To

perform a sensitivity analysis we will change the prior probability distributions of

certain nodes by a 100% range (+ 50% to -50%) in order to test the model‘s

sensitivity to these alternations.



In chapter 4, we discussed in depth the ship specific fatigue factors in order to present

the situation from the naval architect/ designer‘s view. It is undeniable that, most of

the times fatigue aboard ships is the result of the mariner‘s personal decisions. For

example, STCW proposes 8 hours of uninterrupted sleep within twenty four hours; if

a mariner has the chance to sleep but decides to spend this time on a different activity

then the possible accumulation of fatigue would be primarily the result of his own

personal choice.

A naval architect has little jurisdiction over the crew‘s actions during a trip but has the

opportunity to provide a healthy, non-hazardous environment according to the

principles of Human Factors and Applied Ergonomics. After a thorough search on

existing literature and our own subjective knowledge we decided to focus on the

following parameters which we feel can be managed and improved by proper design

in accordance to the science of Naval Architecture and Marine Engineering:

1.Noise

The presence of noise is a common issue in the Maritime domain, especially for those

working in a large engine room where crew members can hardly communicate with

each other. Guidelines and regulations about noise tend to neglect its impact on

fatigue inducing since they focus on the prevention of potential hearing loss.

Temporary loss of hearing can be the result of even short-term exposure to noise and

can eventually lead to permanent hearing loss. Moreover, guidelines about noise fail

to address important physiological outcomes that may appear years after working in a

noisy workplace. Our goal as engineers is to find ways to reduce the noise levels

aboard the ship and isolate its prevalence.

2. Light

Light is an integral part of an ergonomic environment. While the importance of light

is rather obvious the importance of light as far as visibility is concerned, its primary

purpose is to maintain the regular pace of the Circadian Rhythms. Solar light is

essential in calibrating our biological clock making the exposure to natural daylight a

great need. Most mariners have to work night shifts while other crew members

perform their tasks in places where no sunlight illumination is available even at

daytime.( e.g. engine room).

3. Vessel Motions

Working in a constantly moving environment can often induce muscle fatigue since

the vessel‘s motions would probably raise the average energy expenditure of the crew

members. This can be easily explained by the need of a crew member to maintain his

position/stance while performing a specific task at the same time. When the vessel

http://dictionary.in.gr/?searchText=jurisdiction



experiences extreme motions due to severe weather phenomena, the crew member has

to work and rest in an extremely hazardous and fatigue inducing environment. The

Naval Architect can enhance the vessel‘s endurance to these situations by applying

hydrostatic and hydrodynamics principles.

4. Vibrations

According to Calhoun [14], mariners experience shipboard vibrations caused by

machinery, marine equipment and the ship‘s response to the environment. Vibrations

resonate throughout the hull structure and the entire crew can be affected. The

propagation of these vibrations along the decks and bulkheads subject the crew to

whole body vibration and noise. Short-term exposure can lead to headaches, stress,

and fatigue. Long-term exposure leads to hearing loss and causes constant body

agitation. Maritime vibration guidelines keep levels low enough to prevent bodily

injury but the recommended levels can cause fatigue and disrupt sleeping patterns.

5. Management demands

We decided to include this node in our sensitivity analysis as it is a common trend for

engineers nowadays to participate in transportation and operations research

procedures and even receive managerial posts in shipping companies. At this point

our definition of management demands includes arranging shifts and routing, dealing

with the optimization of maritime transport (e.g. classification societies) and the

compliance of legislation on company policies (e.g. ISM code).

8.2.2 SENSITIVITY ANALYSIS

In this paragraph we are going to perform a sensitivity analysis for each of the

parameters discussed in the preceding conversation by changing the prior probability

distribution for a 100% range.

1. NOISE

Table 8. 1: Noise sensitivity analysis

State Prior Probability Distribution Posterior Probability for Fatigue C (%)7

-50% 0.075 0.775 0.25

0 0.150 0.777 0.00

+50% 0.225 0.780 0.38

W.S.8 1.000 0.808 3.90

7 % change of the posterior probability

8 This state describes the worst possible scenario for Noise. We include this option to show what would

be the highest possible value of the posterior probability for this evidence only.



Figure 8. 1: Noise sensitivity analysis

The figure above shows that our model is rather insensitive to the noise parameter.

Since we used estimation when setting the prior distribution for this node, the

outcome of the sensitivity analysis serves to the model‘s reliability and ‗real word‘

reflection. Even though intensive noises can disrupt sleeping patterns severely and

create a fatigue inducing workplace, this evidence cannot raise the posterior

probability by itself.

2. LIGHT

Table 8. 2: Light sensitivity analysis

State Prior Probability Distribution Posterior Probability for Fatigue C (%)

-50% 0.100 0.770 0.90

0 0.200 0.777 0.00

+50% 0.300 0.785 1.01

W.S. 1.000 0.835 7.46

0,772

0,773

0,774

0,775

0,776

0,777

0,778

0,779

0,78

0,781

0,075 0,15 0,225

P(

Fa

tig

ue=

yes

)

P (Noise= extreme) prior probability

FATIGUE-NOISE



Figure 8. 2: Light sensitivity analysis

Contrary to noise, light seems to affect the posterior probability in a more profound

way, a fact that is also shown by the 7.46% increase when poor light is imported as

evidence. We can reasonably assume that this is both the result of a poor working

environment as well as the disruption of the circadian rhythms.

3. VESSEL MOTIONS

Note: Because the Vessels Motions node is a child to the Weather node, we imported

information by removing the parent node in order to assign prior probabilities to this

node.

It is obvious that our model is more sensitive to this parameter than any other

discussed in this chapter. This fact works to our advantage as it was anticipated due to

the vessel‘s motions effect on both working and sleeping environment. To conclude,

this is the field where the naval architect should focus in order to provide a non-

fatigue inducing workplace. The 8.10% overall change is an indicator of this

conclusion.

Table 8. 3: Vessel Motions sensitivity analysis


-50% 0.105 0.769 1.04

0 0.210 0.777 0.00

+50% 0.315 0.786 1.15

W.S. 1.000 0.840 8.10

0,76

0,765

0,77

0,775

0,78

0,785

0,79

0,1 0,2 0,3

P (

Fa

tig

ue=

yes

)

P(Light= Poor) prior probability

Fatigue-Light



Figure 8. 3: Vessels Motions sensitivity analysis

4. VIBRATIONS

Table 8. 4: Vibrations sensitivity analysis


-50% 0.025 0,776 0.128

0 0.050 0,777 0.00

+50% 0.075 0,778 0.127

W.S. 1.000 0,822 5.70

Our previous discussion concerning the noise node applies for vibrations as well. The

model‘s insensitivity to this parameter change proves that estimation for the prior

probability distribution is preferred compared to that of an actual measurement.

Finally, it is essential to point out that since the actual measurement for these

parameters is difficult and often imprecise, the sensitivity analysis vindicates our

choosing Bayesian Network to deal with this uncertainty problem.

0,76

0,765

0,77

0,775

0,78

0,785

0,79

0,105 0,21 0,315

P(F

ati

gu

e=y

es)

P( Vessel Motions= extreme) prior probability

Fatigue-Vessel Motions



Figure 8. 4: Vibrations sensitivity analysis

5. MANAGEMENT DEMANDS

Table 8. 5: Management Demands sensitivity analysis


-50% 0.300 0.753 3.18

0 0.600 0.777 0.00

+50% 0.900 0.801 3.08

W.S. 1.000 0.809 4.11

It was expected that our model would be sensitive to Management Demands for the

following reasons:

The prior probabilities distribution was already high due to the fact that the

maritime industry is a highly demanding domain.

Management Demands affect directly or indirectly workload, sleeping hours,

work condition and overall living quality onboard.

0,775

0,7755

0,776

0,7765

0,777

0,7775

0,778

0,7785

0,025 0,05 0,075

P (

Fa

tig

ue=

yes

)

P(Vibrations= exteme) prior probability

Fatigue-Vibrations



Figure 8. 5: Management Demands sensitivity analysis

8.2.3 NAVAL ARCHITECT OPTIMAL AND WORST SCENARIO

In this paragraph we will present two scenarios in similar approach to that of Chapter

7. The nodes where evidence will be imported are those discussed in our sensitivity

analysis. The order of the information inserted in our network is not random; evidence

was imported according to the following rule:

Evidence was imported in weighing order. The first evidence is the factor that

the sensitivity analysis proved that affects the posterior probability less than

the other nodes investigated.

The actual sequence that we followed is listed below:

Noise (e1)

Management Demands (e2)

Vibrations (e3)

Light (e4)

Vessel Motions (e5)

8.2.3.1 OPTIMAL SCENARIO

Noise (e1) : Normal

Management Demands (e2) : Normal

Vibrations (e3) : Normal

Light (e4) : Sufficient

0,75

0,76

0,77

0,78

0,79

0,8

0,81

0,3 0,6 0,9

P(

Fa

tig

ue=

yes

)

P( Management Demands= demanding) prior probability

Fatigue-Management Demands



Vessel Motions (e5) : Normal

Table 8. 6: Naval Architect Optimal scenario

Evidence 1.Fatigue Posterior Probability P( Fatigue = yes) Change

(%)

No evidence 0,777 0,00

e1 0,772 0,65

e2 0,722 6,93

e3 0,719 0,42

e4 0,700 2,71

e5 0,674 3,86

It is made clear in the following graph that by improving shift rotation, work

schedules and reducing the huge amount of paperwork that a mariner must withstand

while onboard, fatigue levels could drop more in comparison to the rest of the factors

discussed in our sensitivity analysis. Finally, vessel motion seems to come second in

importance.

Figure 8. 6: Naval Architect Optimal scenario

0,65

0,67

0,69

0,71

0,73

0,75

0,77

0,79


P (

Fa

tig

ue=

yes

)

Evidence given

Optimal Scenario



8.2.3.2 WORST SCENARIO

Noise (e1) : Extreme

Management Demands (e2) : Demanding

Vibrations (e3) : Extreme

Light (e4) : Poor

Vessel Motions (e5) : Extreme

Table 8. 7: Naval Architect worst scenario

Evidence 1.Fatigue Posterior Probability P( Fatigue = yes) Change (%)

No evidence 0,777 0,00

e1 0,808 3,99

e2 0,836 3,47

e3 0,856 2,39

e4 0,881 2,92

e5 0,891 1,14

Figure 8. 7: Naval Architect worst scenario

0,77

0,79

0,81

0,83

0,85

0,87

0,89

0,91


P(F

ati

gu

e =

yes

)

Evidence given

Worst Scenario



Contrary to the optimal scenario, we can assume by the form of the figure, that all

factors affect the posterior fatigue probability in a much similar way. The overall %

change drops after the import of each evidence thus proving that when the situation is

already dangerous it is difficult to worsen more.

Figure 8. 8: Naval Architect scenario comparison

0,65

0,7

0,75

0,8

0,85

0,9

0,95

no

evidence

e1 e2 e3 e4 e5

P(F

ati

gu

e =

yes

)

Naval Architect scenario

comparison

Optimal sc.

Worst sc.



CHAPTER 9 CONCLUSIONS



9 CONCLUSIONS

9.1 GENERAL LITERATURE

In this paragraph we will list our general conclusions that are the product of our

extended research upon existing literature. The most important presumption can be

found below:

Even though Human Factors are responsible for almost 80% of maritime accidents

and casualties, our industry is way behind in terms of research and ergonomics

implementation compared to more advanced industries such as aviation and

railroad. While constant effort is made towards proper integration of ergonomic

principles on ships, the maritime domain has a long road ahead in order to fully

benefit from Human Factors. Consequently, fatigue is under researched as part of

HF science while it also underrated compared to other ergonomic issues such as

human-machine interaction. We can only explain this tendency as a result of

constant economic pressures and the need of shipping companies to reduce

manning levels and operational costs. An interesting finding is that the United

States Coast Guard, a nonprofit government organization, is a pioneer in the field

of Human Factors engineering and fatigue mechanisms understanding and

countermeasures in particular. This fact leads to the conclusion that fatigue should

be properly researched firstly from the mariner‘s perspective and then of that of

the shipping company in order to provide beneficial solutions for both parties.

IMO has provided guidelines regarding fatigue mitigation aboard ship but the

majority of scholars consider these propositions outdated. It took more than fifteen

years to revise STCW, fifteen years where manning levels became even lower and

the maritime domain became even more competitive and demanding. Most critics

agree IMO‘s existing approaches seem largely inadequate while their

improvement is clearly regarded as one strategy that could reduce the problem.

IMO has the means and the authority to make fatigue awareness mandatory for

mariners and shipping companies. The introduction of the Human Element

working group and its interplay with various maritime organizations and parties

(e.g. the FSA group) has provided an improved framework in order to properly

eliminate fatigue-related hazardous outcomes.

It is very common to regard fatigue as the main factor behind impaired

performance and reduced safety aboard ships. While this is undeniably true,

accumulative fatigue can have long term consequence that few mariners are aware

of. A common example is the exposure to low density noises or vibrations which

are not regarded as a potential threat in the workplace. However, scientific

research has shown that constant exposure to vibrations and noises, even if their



density does not seem threatening, can result to chronic fatigue and other long

term deceases (e.g. spastic colitis). At this point, it is essential to indicate that

almost every guideline or regulation addresses only short term issues while long

term outcomes are completely neglected. In order to understand the preceding

discussion we will provide a representative example of a common incident aboard

a merchant vessel:

The regulation regarding permitted noise levels in the engine room were compiled

in order to prevent permanent loss of hearing and difficulties in communication

amongst crew members. However, studies have shown that exposure to noises of

half the density proposed in the regulations is sufficient enough to induce fatigue

and result in medical problems that may appear long after the mariners years of

service. The question remains: How many mariners are aware of this fact and

how many owners can understand that even these seemingly minor problems can

result in casualties and compensations?

Most surveys regarding mariners‘ fatigue have proved that the vessel‘s type and

scheduling is a critical factor regarding fatigue. Questionnaires‘ findings show

that crew member working on mini bulkers or small general cargo vessels tend to

get tired more quickly than those who serve on large oil tankers. A logical

explanation for these findings is the fact that the workload on short sea shipping

vessels is heavier due to the regularity of port calls and the difficulty to adjust the

mariner‘s lifestyle to the short trip-land-short trip sequence. Therefore, it is easier

for a seafarer working on VLCC to adjust his daily routine in a 20 year trip than a

crew member working on a 5000 DWT general cargo vessel that may have 10 port

calls within a calendar month. It is crucial to mention that whenever a ship is at

port work shift routines are often abolished while crew members rest for very

small periods thus resulting in sleep deprivation and circadian disruption. Finally,

we must add that mariners working on containerships and RO/RO carriers,

independently of size, also experience similar issues due to the nature of the

vessel‘s trading schedules.

The implementation of advanced technology aboard ships has resulted in the

constant decrease of manning levels while the need to cut operating costs has

provided as a fertile field for this practice. The majority of mariners complain

about working hours and insufficient time for rest or the poor quality of onboard

recreational activities. A typical example of this practice would be the disaster of

Exxon Valdez .Whereas tankers in the 1950s carried a crew of 40 to 42 to manage

about 6.3 million gallons of oil; the Exxon Valdez carried a crew of 19 to transport

53 million gallons of oil. In 2010, the complements policies are more flexible than

ever. While certain Flag States try to keep strict policies regarding manning levels,

the majority allows vessels to sail with minimum complements. In addition, the

integration of highly complex and sophisticated machinery aboard ships has



created the need for extensive training while it impaired the mariner‘s initiative

and understanding of his surroundings.

The issue that we feel is of the outmost importance when addressing fatigue, is

the proper knowledge and understanding of the subject. Most people working

aboard merchant ships are unaware of fatigue mechanisms and its causes. While

various organizations (e.g. STCW and Classification societies) have tried to

provide sufficient information about fatigue, sleeping patterns and circadian

rhythms, questionnaire findings and several other interviews indicate that

seafarers‘ overestimate their capabilities and exceed theirlimitations thus making

themselves vulnerable to fatigue prevalence. Whether they knew what they were

up against, they would probably think twice before endangering their personal

safety and the ship‘s as well. Where legislation and guidelines focus on what

might cause fatigue, they rarely mention the effects on general well being and

performance. Finally, since fatigue cannot be measured precisely, the importance

of proper fatigue perception is integral for an individual to understand when his

limitations have been surpassed and further effort will eventually result in

underperformance and dangerous incidents.

Recent statistical data show that fatigue related incidents are almost 20% of all

maritime accidents. We think that this percentage is an underestimation since most

errors recorded, even if there is no direct connection to fatigue, are often result of

a tired individual. For example, in a conversation between two sleep deprived

officers, if there is a misunderstanding in a specific command, the error will be

recorded as a communication error. Knowing that fatigue can impair an

individual‘s hearing ability, this error now seems more of a fatigue error than a

communicational one. Therefore, most reports fail to recognize this fact thus

neglecting fatigue‘s actual importance. In addition to the preceding conclusion,

since most mariners cannot recognize the actual effects of fatigue, it is logical that

many slips and errors are recorded in a different manner. Our rough estimation is

that fatigue related accidents are over 50% of the overall maritime incidents given

the fact that the maritime industry is often connected to heavy workloads, pressing

deadlines and other demands.

Furthermore, it is rather profound that there is difficulty in understanding and

implementing fatigue related human factors in ship design. Naval architects focus

on the vessels endurance to waves, its overall stability and the optimization of its

service speed. When it comes to designing the superstructure, their top priority is

to ensure its structural resilience and overall interaction with the hull. However,

the proper designing of the sleeping quarters and recreational facilities, naval

architects neglect the importance of the working and sleeping environment in

order to reduce costs and enhance the vessel‘s performance at sea. For example

the fact that the sleeping quarters are just above the engine room can prove a risk

factor for fatigue. The engine room is the source of extreme noises, vibrations and



even high temperatures and humidity which, with the lack of proper dampening,

make their way to the superstructure and into the sleeping quarters creating an

unhealthy sleeping/living environment for crew members. Finally, few studies

have highlighted the impact of the vessel‘s motions on the mariner‘s performance

and the difficulties that may come with working in a constantly moving

environment.

Proper lighting aboard ships has been highlighted as a major issue regarding

fatigue, lack of alertness and the circadian rhythms. Most guidelines focus on

visibility issues while the impact of poor lighting on the disruption of the

circadian rhythms pattern is addressed with desultoriness. Crew members working

in the engine room, even during daytime, lack the salutary effect of sunlight.

Watchkeepers who perform night shifts in a regular basis tend to have alertness

issues due to their biological clock malfunctioning. In conclusion, lighting quality

is not yet regarded as major contributing fatigue risk factor by IMO or the ISM

code. Nevertheless, our thesis, in accordance to the existing literature, pinpoints

the need for improvement and more thorough revision on existing legislation.

9.2 CONCLUSION FROM OUR BAYESIAN NETWORK APPLICATION ON

MARITIME SCENARIOS

The following conclusions are products of our Bayesian Network application analysis

on real and hypothetical scenarios:

Note: Some of the findings were general and therefore were included in the previous

paragraph (e.g. vessels motions)

Our major finding was that human physiology (sleep and circadian rhythms) is

the milestone in any discussion or study regarding fatigue. The safeguard of

sleeping quality and the proper function of the biological clock is the integral

step towards fatigue mitigation. Even if all other factors such as environment

and work condition are favorable, poor sleeping quality and drowsiness will

always result in fatigued individuals and underperformance. Taking this

general rule into consideration, we realize that our best chance to enhance

sleep quality is by providing better work schedules, sufficient rest

opportunities and heavy workload discharging. As far as the naval architect‘s

perspective is concerned, there is a great potential for improvement in both

working and sleeping conditions, the most prominent being the opportunity to

ameliorate the connectedness between the vessel‘s motion and limitations (e.g.

difficulty to sleep in a moving environment).

All examples investigated proved to be sensitive to one specific parameter:

the monotonous nature of watchkeeping. While examining the figures

provided by each scenario, one thing was made perfectly clear; tedious jobs



can radically increase the fatigue posterior probability. The situation could

easily get out of hand when working night shifts. On the other hand, any

attempts made to provide a more interesting kind of task (e.g. Betty James)

often result in distracting environments. Finally, even if most concerned

parties agree that the nature of that task can hardly change, we believe that it is

of the outmost importance to provide a less tedious task when it comes to

watchkeeping.

Another interesting fact that came to our knowledge is that mariners often lie

on official documents about rest times. These facts often come into

prominence after a major disaster, when investigations find serious

contradictions between official documents and mariners‘ actual testimonies.

This practice is a typical procedure of underreporting/misreporting which

allows mariners to infringe policies and regulations. This also confirms the

fact we discussed earlier about the seafarers‘ ignorance and poor knowledge

regarding fatigue mechanisms and its possible outcomes. Our main

assumption about this practice is the following: The seafarers are risking much

more than they wish to avoid.

Most of the accidents related to fatigue involve a sole watchkeeper on the

bridge. Flag state policies and other legislation make the presence of at least

two watchkeepers on the bridge mandatory at all times. Unfortunately, the

tendency to reduce operating cost has led shipping companies to reduce

manning levels as well, making this situation unavoidable. After many years

of debating, most parties agree that extra watchkeepers should be made

mandatory in order to avoid further disasters in the maritime domain.

At this point we must indicate that our graphical model was generally

compiled having indoors activities in mind. Our Bayesian Network makes the

assumption that most crew members work indoors so they are not exposed to

weather and temperature. In a real time scenario, this could not be true as

many duties on-board require outdoor activities (e.g. small fishing vessels).

Therefore, it would be correct to assume that the major findings of this thesis

are addressed to mariner‘s working mostly indoors (especially officers and

engineers). However, our general approach to fatigue and our thorough

description of its consequences could serve as useful instrument for any

mariner serving on a merchant vessel.

While most seafarers believe that their training is sufficient to overcome any

obstacle during a day at work, our study has found that: a) crew members tend

to overestimate the level of their training and general competence b) older,

more experienced mariners have lesser stress and anxiety problems. Finally,

we must mention that the presence of fatigue and the lack of proper training

can be a rather lethal combination that may result in major disaster.



More and more mariners complain about the quality and availability of

recreational facilities aboard ship. The presence of such installations has

proven to be an effective measure to reduce stress levels and enhance the crew

members‘ morale. Furthermore, internet availability is essential, especially for

those serving on deep sea shipping vessels that spend months away from

families and friends.

9.3 SUGGESTIONS

In this paragraph we will present our suggestion for each of the conclusions discussed

in this paragraph and several other methods that may proven to be useful for

mitigating fatigue in the maritime domain:

Proper integration of Human Factors Engineering into Maritime Transport and

Ship design

Emphasis on physiology and the Circadian Rhythms through further research

and recognition of fatigue as a primary factor in maritime casualties

Research from the seafarer‘s perspective instead of the shipping company‘s

Further improvement and revision of the Guidelines proposed by IMO and the

amendments made by STCW with emphasis on working/rest hours

Stricter Flag State legislation regarding manning and watchkeeping

Better understanding of the long term effects of fatigue and exposure to low

density noises and vibration. Assistance from the medical field

Revision of Guidelines and regulations regarding habitability with specific

emphasis on permitted noise, vibrations, humidity levels and lighting quality

Higher manning levels, especially for short sea shipping vessels and

containerships. Port assistance during discharging (especially for large tankerd

and containerships)

Recognition of the fact that a human could not be replaced by any machine or

software

Proper education about fatigue in general. Crew seminars with emphasis on

fatigue prevalence, personal choices and countermeasures. Identify fatigue

perception as a major concern



Improved in-depth techniques for categorization of human error in order to

highlight the importance of fatigue at sea

The implementation of fatigue related issues into the principles of Naval

Architecture and Marine Engineering. The naval architect should not neglect

habitability issues in favor of the vessel‘s performance. Both aspects should be

treated with similar interest and graveness

Installation of more efficient lighting systems with emphasis on the Circadian

Rhythms mechanism

Improved, ergonomic sleeping quarters. Sleep is fundamental when discussing

fatigue so there should be constant effort for improving sleeping environments

and the overall sleeping quality

The watchkeeper‘s task has to become less monotonous and passive while two

crew members must remain on the bridge at all times. Legislation should be

even stricter regarding this matter

Countermeasures for underreporting errors and non-compliance to regulations

through more regular inspections and thorough investigations of official

documents regarding rest hours and work schedules

Introduction of improved and ergonomic recreational facilities. Sufficient

spaces for accommodation for all crew members and implementation of

contemporary facilities such as internet connection and DVD‘s

Proper training with emphasis in emergency situations and human-machine

interaction through simulation procedures

Last but not least, fatigue can be eliminated by adjusting to a healthy life style.

Personal decisions and situation awareness is very important in order not to

overestimate the human body‘s capabilities and exceed its limitation. Mariners

should understand that by not reporting fatigue or by infringing regulations

regarding rest hours, they are putting themselves and the ship in great danger

while embracing a hazardous practice that has led the maritime industry into

major catastrophes, fatalities and large ecological disasters.



REFERENCES

BOOKS, PAPERS & REPORTS

1) Aahsberg E. (1998). Perceived fatigue related to work. Internal report, University

of Stockholm, Department of Psychology, Sweden. ISBN 91-7153-830-5.

2) Adhikari S., Bayley C., Bedford T., Busby J., Cliffe A., Devgun G., Eid M.,

French S., Ri Keshvala R., Pollard S., Soane E., Tracy D., Wu S. (2009).

Human Reliability Analysis: A Review and Critique. Manchester Business

School Working Paper No 589.

3) Akerstedt T. (2000). Consensus statement: Fatigue and accidents in transport

operations. Journal of Sleep Research, 9, 395.

4) Akerstedt T. (1998). Shift work and disturbed sleep/wakefulness. Sleep Medicine

Reviews, Vol. 2, No. 2, pp 117-128, 1998.

5) Åkerstedt T. (2006). Sleepiness and Circadian Rhythm Sleep Disorders. Sleep

Med Clin 1 (2006) 17–30.

6) Akerstedt T., Landstrom U. (1995). Work place countermeasures of night shift

fatigue. International Journal of Industrial Ergonomics 21 1998 167-178.

7) Ben-Gal I. (2007). Bayesian Networks.In Ruggeri F., Faltin F. & Kenett R.,

Encyclopedia of Statistics in Quality & Reliability, Wiley & Sons (2007).

8) Bobbio A., Portinale L., Minichino M., Ciancamerla E. (2001). Improving the

analysis of dependable systems by mapping fault trees into Bayesian networks.

Reliability Engineering and System Safety 71 (2001) 249–260.

9) Boverie S., (2004). Driver Fatigue Monitoring Technologies and future ideas.

Road safety workshop 2004, Balocco, Italy.

10) Breieriova L., Choudhari M., (1996). An Introduction to Sensitivity Analysis. MIT

System Dynamics in Education Project

11) Bridges B., Johansson P., Pearson L. Using Risk Engineering Grading Systems to

assess Organizational Fatigue Risk.

12) Caldwell J.A. (2001). The Impact of Fatigue in Air Medical and Other Types of

Operations: A Review of Fatigue Facts and Potential Countermeasures. Air

Medical Journal 20:1.



13) Caldwell J.A., Caldwell J.L., Schmidt R.M., (2008). Alertness management

strategies for operational contexts. Sleep Medicine Reviews (2008) 12, 257–273.

14) Calhoun S.R., (2006). Human Factors in Ship Design: Preventing and Reducing

Shipboard Operator Fatigue. University of Michigan Department of Naval

Architecture and Marine Engineering U.S. Coast Guard Research Project.

15) Celik M., Cebi S. (2009). Analytical HFACS for investigating human errors in

shipping accidents. Accident Analysis and Prevention 41 (2009) 66–75.

16) Chang Y.H.J., Mosleh A., (2006) Cognitive modeling and dynamic probabilistic

simulation of operating crew response to complex system accidents. Part 2:IDAC

performance influencing factors model. Reliability Engineering and System

Safety 92 (2007) 1014–1040.

17) Charniak E. (1991). Bayesian Networks without Tears. AI magazine.

18) ClassNK (2010). Guidelines for the prevention of human error aboard ships.

19) Cook T.C., Shipley P. (1980). Human factors studies of the working hours of

UK‟ship pilots. Applied Ergonomics 1980, 11.2, 85-92

20) Deacon T., Amyotte P.R, Khan F.I., (2010). Human error risk analysis in offshore

emergencies. Safety Science 48 (2010) 803–818.

21) Dhillon B.S., Rayapati S.N., (1987). Human Performance Reliability Modelling.

Pergamon 0026-2714/8853.00

22) DiMattia D.G., Khan F.I, Amyotte P.R., (2005). Determination of Human Error

Probabilities for Offshore Platform Musters. Journal of Loss Prevention in the

Process industries Volume 18, Issues 4-6, July-November 2005, Pages 488-501.

23) Dobie T.G. The Importance of the Human Element in Ship Design. National

Biodynamics Laboratory, College of Engineering, University of New Orleans,

New Orleans.

24) Dougherty E. M. Jr. (1990). Human reliability analysis - Where should thou turn?

Reliability Engineering and System Safety, 29(3), 283-299.

25) Eastridge B.J., Hamilton E.C., O‘Keefe G.E., Rege R.V., Valentine R.J., Jones

D.J., Tesfay S., Thal E.R., (2002). Effect of sleep deprivation on the performance

of simulated laparoscopic surgical skill. The American Journal of Surgery 186

(2003) 169–174.

http://www.sciencedirect.com/science/journal/09504230

http://www.sciencedirect.com/science/journal/09504230

http://www.sciencedirect.com/science?_ob=PublicationURL&_tockey=%23TOC%235255%232005%23999819995%23606759%23FLA%23&_cdi=5255&_pubType=J&view=c&_auth=y&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=a4d69a30c23580e74b9003ee6fd3d590



26) Folkard S., Åkerstedt T. (2004). Trends in The Risk of Accidents and Injuries and

Their Implications for Models of Fatigue and Performance. Aviation, Space, and

Environmental Medicine, Vol. 75, No. 3, Section II, March 2004.

27) Gander, P. (2005). A review of fatigue management in the maritime sector.

Massey University Sleep/Wake research centre.

28) Hancock P.A., Desmond P.A. (2001). Stress, Workload and Fatigue. Mahwah,

NJ: Lawrence Erlbaum.

29) Harati-Mokhtari A., Wall A., Brooks P., Wang J. Automatic Identification System

(AIS): A Human Factors Approach. Liverpool John Moores University

30) Heckerman D., (1995). A Tutorial on Learning with Bayesian Networks.

Technical Report MSR-TR-95-06.

31) Hetherington C., Flin R. & Mearns K. (2006). Safety in shipping: The human

element. Journal of Safety Research, 37 (4): 401-411.

32) Hollnagel E. (1998). Cognitive reliability and error analysis method. Oxford:

Elsevier Science Ltd.

33) Houtman I., Miedema M., Jettinghoff K., Starren A.,Heinrich J., Gort J. (2005).

Fatigue in the shipping industry. TNO Report 20834/11353

34) Hoeyer J.L, Hansen H.L., (2005). Obesity amongst Danish Seafarers. Int Marit

Health. 2005;56(1-4):48-55.

35) HSE (2006). The Fatigue and Risk Index Calculator. HSE Research Report 446

(2006).

36) ΙΜΟ (2001). Guidance in Fatigue Mitigation and Management.

37) IMO (2002). Guidelines for the application of Formal Safety Assessment (FSA)

for use in the IMO rule-making process.

38) ILO Convention No. 180 concerning Seafarers' Hours of Work and the Manning

of Ships. Geneva, 22 October 1996.

39) International Transport Workers‘ Federation, „Fight fatigue‟.

40) Jensen F.V. (1996). An introduction to Bayesian networks. UCL Press, London,

UK.

http://www.ncbi.nlm.nih.gov/pubmed?term=%22Hoeyer%20JL%22%5BAuthor%5D

http://www.ncbi.nlm.nih.gov/pubmed?term=%22Hansen%20HL%22%5BAuthor%5D

javascript:AL_get(this,%20'jour',%20'Int%20Marit%20Health.');

javascript:AL_get(this,%20'jour',%20'Int%20Marit%20Health.');



41) Jensen F.V., (2001). Bayesian networks and decision graphs. Springer-Verlag,

New York (USA).

42) Ji Q., Lan P., Looney C. (2005). A Probabilistic Framework for Modeling and

Real Time Monitoring Human Fatigue. University of Nevada, Reno.

43) Ji Q., Lan P., Looney C. Information Fusion with Bayesian Networks for

Monitoring Human Fatigue .University of Nevada, Reno.

44) Jones C.B, Dorrian J., Rajaratnam S.M.W., Dawson D. (2005). Working hours

regulations and fatigue in transportation: A comparative analysis. Safety Science

43 (2005) 225–252.

45) Kato Y., Endo H., Kizuka T. (2008). Mental fatigue and impaired response

processes: Event-related brain potentials in a Go/NoGo task. International Journal

of Psychophysiology 72 (2009) 204–211.

46) Khan F.I., Amyotte P.R., DiMattia D.G., (2005). HEPI: A new tool for human

error probability calculation for offshore operation. Safety Science 44 (2006)

313–334.

47) Kahn-Greene E.T., Killgore D.B., Kamimori J.H., Balkin T.J., Killgore W.D.S.

(2006). The effects of sleep deprivation on Symptoms of psychopathology in

healthy adults. Sleep Medicine 8 (2007) 215–221.

48) Kirwan B., Kennedy R., Taylor-Adams S., and Lambert B. (1995). The validation

of three Human Reliability Quantification techniques - THERP, HEART and

JHEDI: Part II - Results of validation exercise. Applied Ergonomics Vol 2X. No.

I. pp. 17-25 1997

49) Kirwan B., Ainsworth L. (Eds.) (1992). A guide to task analysis. Taylor & Francis

50) Learmount D. (2009). Proof that fatigue kills. FRMS Forum, Qinetiq Farnborough

51) Leung A.W.S., Chan C.C.H, Wong P.C.C, Ng J.J.M, (2005). Factors contributing

to officers‟ fatigue in high-speed maritime craft operations. Applied Ergonomics

37 (2006) 565–576.

52) Lutzhoft M., Thorslund V., Kircher A., Killberg M., (2007). Fatigue at sea: A

field study on Swedish shipping. VTI rapport 586A.

53) Macrae C., (2009). Human factors at sea: common patterns of error in

groundings and collisions. Maritime Policy & Management, 36: 1, 21- 38



54) Marine Accident Investigation Branch (MAIB). (2004). Bridge Watchkeeping

Safety Study. Entre (S.I.R.C).

55) Marine Accident Investigation Branch (MAIB), (2009). Antari grounding report.

No 7/2009.

56) Marine Accident Investigation Branch (MAIB), (2000). MFV Betty James. Report

No 34/2000.

57) Marine Accident Investigation Branch (MAIB), (2002). Report on the analysis of

Fishing Vessel accident data 1992 to 2000.

58) McCafferty D.B., Baker C.C, (2002). Human Error and Marine Systems: Current

Trends. IBC 2nd Annual Conference on Human Error, London, England.

59) McCallum M., Raby M., Rothblum A. (1996). Procedures for Investigating and

Reporting Human Factors and Fatigue Contributions to Marine Casualties.

USCG Report No CG-D-09-97.

60) McSweeney K., Pray J., Craig B.N. (2009). Integration of Human Factors

Engineering into Design- an applied Approach. ABS Technical Papers 2009.

61) McSweeney, K., (ABS). Crew Habitability: What You Need to Know.

62) Mital A., Bishu R.R., Manjunath S.G. (1991). Review and evaluation of

techniques for determining fatigue allowances. International Journal of Industrial

Ergonomics, 8 (1991) 165-178.

63) Moore A.W. (2001). Bayes Nets for representing and reasoning about

Uncertainty. School of Computer Science, Carnegie Mellon University.

64) Moore-Ede M., Heitmann H., Guttkuhn R., Trutschel U., Aguirre A., Croke D.,

(2004). Circadian Alertness Simulator for Fatigue Risk Assessment in

Transportation: Application to Reduce Frequency and Severity of Truck

Accidents. Aviation, Space, and Environmental Medicine Vol. 75, No. 3, Section

II , March 2004.

65) Moore W.H. (1992). The Grounding of Exxon Valdez: An examination of the

Human and Organizational Factors. Society of Naval Architects and Marine

Engineers, Northern California Section, Student Paper Presentation, 1-15.

66) Moore W.H., Bea R.G. (1992). Modeling the effects of Human Error from Post-

Mortem Marine Casualties Studies. Report HO-92-4, University of California,

Berkeley.

http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=ADA323392

http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=ADA323392



67) Murphy K. (2001). A Brief Introduction to Graphical Models and Bayesian

Networks. UC Berkeley,2001.

68) Nillson T., Nelson T.M., Carlson D. Development of fatigue symptoms during

simulated driving. Pergamon PII: SOOOl4575 (97)00027-4.

69) Onisko A., Druzdzel M.J., Wasyluk H. (2000). Learning Bayesian Network

Parameters from Small Data Sets: Application of Noisy-OR Gates. 12th European

Conference on Artificial Intelligence (ECAI-2000), Berlin, Germany, 22 August

2000.

70) ORR (2006). Managing fatigue in safety critical work Railways and Other Guided

Transport Systems. Safety Regulations 2006.

71) Parker A.W. , Balanda K., Briggs L., Hubinger L.M. A survey of the work and

sleep patterns of Great Barrier Reef Pilots. AMSA, ISBN: 0-642-47618-7.

72) Park K.S., Jung K.T., (1995). Considering performance shaping factors in

situation-specific human error probabilities. International Journal of Industrial

Ergonomics 18 (1996) 325-331

73) Pearl J. (1988). Probabilistic Reasoning in Intelligent Systems: Networks of

Plausible Inference. Morgan Kaufmann Publishers, Inc., San Mateo, CA, 1988.

74) Phillips H. (2000). Sleep, watchkeeping and accidents: a content analysis of

incident at sea reports. Transportation Research Part F 3 (2000) 229-240.

75) Pomeroy R.V., Sherwood Jones B.M. (2002). Managing the Human Element in

Modern Ship Design and Operation. Human Factors in Ship Design and

Operation Conference, October 2002.

76) Psaraftis H.N., Caridis P., Desypris N., Panagakos G, Ventikos N. The Human

Element as a Factor in Marine Accidents. Proceedings of the 1st Greek Maritime

Transport Conference, Piraeus, Greece, pp. 45-53.

77) Psarros G., Skjong R., Eide M.S. (2009). Under-reporting of maritime accidents.

Accident Analysis and Prevention 42 (2010) 619–625.

78) Raslear T.G., Coplen M. (2004). Fatigue Models as Practical Tools: Diagnostic

Accuracy and Decision Thresholds. Aviation, Space, and Environmental Medicine

Vol. 75, No. 3, Section II ,March 2004.

79) Rosqvist T., (2005). Fatigue Analysis for Fleet Management Using Bayesian

Networks. VTT Working Papers 35.

http://linkinghub.elsevier.com/retrieve/pii/S1369847801000079

http://linkinghub.elsevier.com/retrieve/pii/S1369847801000079



80) Rothblum A.M. Human Error and Marine Safety. U.S. Coast Guard Research &

Development Center.

81) Ruggeri F., Pedrali M., Trucco P., Cagno E., (2004). On the use of Bayesian

Belief Networks in modelling human and organizational causes of maritime

accidents.ENBIS, Copenhagen, Denmark

82) Sanquist T.F., Lee J.D., (1993). Human Factors Plan for Maritime Safety:

Annotated Bibliography. Report No.CO-D-08-93, National Technical Information

Service.

83) Sanquist T., Raby M. (1996). Fatigue and Alertness in Merchant Marine

Personnel: A Field Study of Work and Sleep Patterns. Washington, D.C.: Battelle

Seattle Research Center.

84) Sanquist T.F., Raby M., Forsythe A., Carvalhais A.B. (1997). Work hours, sleep

patterns and fatigue among merchant marine personnel. European Sleep Research

Society, J. Sleep Rex, 6, 245-251.

85) Scott J.P.R., McNaughton L.R., Polman R.J.C., (2004). Effects of sleep

deprivation and exercise on cognitive, motor performance and mood. Physiology

& Behavior 87 (2006) 396 – 408.

86) Shen J., Barbera J., Shapiro C.M., (2005). Distinguishing sleepiness and fatigue:

focus on definition and measurement. Sleep Medicine Reviews (2006) 10, 63–76 .

87) Sherry P. (2000). Fatigue Countermeasures in the Railroad Industry: Past and

Current Developments. Association of American Railroads 2000.

88) Sherry P., Philbrick K.E., (2004). The Effects of Individualized Actigraph

Feedback on Fatigue Management in Railroad Engineers. NCIT Final Report.

89) Smith A.P., Allen P.H., Wadsworth E.J. (2007). A comparative approach to

seafarers‟ fatigue. In Proceedings of International Symposium on Maritime

Safety, Security and Environmental Protection. Athens, Greece.

90) Smith A.P., Allen P., Wadsworth E.J., (2006). Seafarers‟ fatigue: the Cardiff

Research Programme. Report to MCA and HSE.

91) Smith A.P., Lane T., Bloor M. Fatigue Offshore: A Comparison of Offshore Oil

Support Shipping and the Offshore Oil Industry. University of Cardiff.



92) Smith A.P., Lane T., Bloor M., Allen P., Burke A., Ellis N. (2003). Fatigue

offshore: Phase 2 the short sea and coastal shipping industry: Seafarers

International Research Center.

93) Smith A.P. (2001). Offshore fatigue: A study of ships in the offshore oil industry.

Paper presented at the SIRC second symposium, Cardiff University.

94) Smith A.P. (2007). Seafarer fatigue where next? International Transport Workers'

Federation (ITF).

95) Smith A.P., (2007). Adequate crewing and seafarers fatigue: The International

perspective. International Transport Federation, London

96) Smith A.P. (2010). Noise and Health: Why we need more research.

INTERNOISE June 2010 Lisbon, Portugal.

97) Srinivas, S. (1992). A generalization of the Noisy-OR model. Proceedings of the

Ninth Annual Conference on Uncertainty in Artificial Intelligence (UAI–93), pp.

208–215, Washington, D.C., (1993).

98) STCW Convention and Code, 2010 Manila Amendments. Manila Philippines

99) Squire D. The Human Element in shipping. Alert! – The International Maritime

Human Element Bulletin.

100) Sugden C., Darzi A. (2010). The performance effects of sleep deprivation and

fatigue. Surgery February 2010 255-256.

101) Sumner W., Zhong Xu J.H. (2002). Modeling Fatigue. AMIA 2002 Annual

Symposium.

102) Trucco P., Cagno E., Ruggeri F., Grande O. (2006). A Bayesian Belief

Network modeling of organizational factors in risk analysis: A case study in

maritime transportation. Reliability Engineering and System Safety 93 (2008)

823–834.

103) USCG (2006) Crew Endurance Management. Commandant Instruction

3500.2.

104) USCG, Marine Safety and Environmental Protection (G-M), (2003). Crew

Endurance Management System (CEMS). Report No. CG-D-01-03

105) USDA Forest Service. Fatigue Awareness. Missoula Technology and

Development Center.



106) Vanderhaege F. (2000). A non-probabilistic prospective and retrospective

human reliability analysis method - application to railway system. Reliability

Engineering and System Safety 71 (2001) 1–13.

107) Van Dongen H. P. A., Dinges D.F., Circadian Rhythms in Fatigue, Alertness

and Performance. Principles and Practice of Sleep Medicine (3rd ed.), pp. 391–

399.

108) Ventikos N., Papamichalis G., (2008). Μελέηη ναςηικών αηςσημάηων μέζω

δικηύων πίζηηρ, ηεύσορ Α: Μονηέλο πποζάπαξηρ. National Technical University of

Athens, (in Greek).

109) Ventikos N.P., Psaraftis H.N. (2003). Spill accident modeling: a critical

survey of the event-decision network in the context of IMO‟s formal safety

assessment. Journal of Hazardous Materials 107 (2004) 59–66.

110) Ventikos N.P., Lyridis D.V., Lykos G.V., Logothetis T. (2010). Identifying

and assesing Non-Technical Skills on Greek Maritime Officers: The story under

the spotlight. HPAS 2010, 16-18 June, Glasgow, Scotland, United Kingdom.

111) Yang G., Lin L., Bhattacharya P. (2006). A driver fatigue recognition model

based on information fusion and dynamic Bayesian network. Information Sciences

180 (2010) 1942–1954.

112) Yousef G.S. (2005). Stress Management. Center for Advancement of Post-

Graduate Studies and Research in Engineering Sciences at Cairo University, ISBN

977-223-988-4.

113) Zhang N.W., Poole D. A simple approach to Bayesian network computations

114) Λάηνο Λ. (2003). ύγσπονη Επγονομία. Παπαζσηεξίνπ ISBN 9607530446.

(in Greek)

115) Ξέλνο Μ. (2003) Μονηελοποίηζη ηηρ ςμπεπιθοπάρ ηων θοιηηηών ηηρ

Πληποθοπικήρ με Χπήζη Bayesian Networks. 2o Παλειιήλην πλέδξην γηα ηελ

Αλνηθηή θαη Δμ Απνζηάζεσο Δθπαίδεπζε, Πάηξα, pp. 318-329, 2003 (in Greek)

http://www.papasotiriou.gr/results.asp?field=Publisher&ty=1&aid=6&av=%CE%A0%CE%B1%CF%80%CE%B1%CF%83%CF%89%CF%84%CE%B7%CF%81%CE%AF%CE%BF%CF%85



WEBSITES

A) AviationKnowledge.wiki.com:http://aviationknowledge.wikidot.com/aviation:

circadian-rhythms

B) ATSB 2010: http://www.abc.net.au/news/stories/2010/04/15/2873531.htm

C) HSE/Human/Factors/Fatigue:http://www.hse.gov.uk/humanfactors/topics/fatig

ue.htm

D) Learning and analyzing Bayesian networks (2007):

http://controls.engin.umich.edu/wiki/index.php/Learning_and_analyzing_Baye

sian_networks_with_Genieetworks_with_Genie

E) MarineBuzz.com(2008) How Seafarers can Overcome Fatigue ?

:http://www.marinebuzz.com/2008/05/07/how-seafarers-can-overcome-

fatigue/

F) PILOTFIREND.COM:http://www.pilotfriend.com/training/flight_training/hu

man/drugs_etc.htm

G) U.S.Department of Transportation, Operator Fatigue Management (OFM)

Program :http://hfcc.dot.gov/ofm/

H) Queensland Government, Workplace Health and Safety topics, Fatigue:

http://www.deir.qld.gov.au/workplace/subjects/fatigue/about/index.htm

I) Wikipedia-Human Factors: http://en.wikipedia.org/wiki/Human_factors

J) Wikipedia-Human Reliability: http://en.wikipedia.org/wiki/Human_reliability

K) IMO-Human Element working group:

http://www.imo.org/humanelement/mainframe.asp?topic_id=286

L) Handicapsupport.com/fatigue:http://handicapsupport.com/Physical-

Disability/Fatigue/

M) Wikipedia- fatigue: http://en.wikipedia.org/wiki/Fatigue_%28medical%29

N) Free-dictionary –fatigue, the Columbia Electronic Encyclopedia® Copyright

© 2007, Columbia University:

http://encyclopedia2.thefreedictionary.com/Fatigue

O) Wikipedia-sleep: http://en.wikipedia.org/wiki/Sleep

P) Healthcentral.com–nap:http://www.healthcentral.com/sleep-

disorders/napping-192940-5.html

Q) Pilot fatigue By Dr. Samuel Strauss:

http://www.pilotfriend.com/aeromed/medical/fatigue1.html

R) AviationKnowledge-circadian-rhythms:

http://aviationknowledge.wikidot.com/aviation:circadian-rhythms

S) NASA Ames Fatigue Countermeasures Program: http://human-

factors.arc.nasa.gov/zteam/fcp/FCP.pubs.html

T) Wikipedia-stress : http://en.wikipedia.org/wiki/Stress_%28biology%29

U) Universite Laval –stress: http://www2.ulaval.ca/en/home.html

V) ITF Seafarer Fatigue: Wake up to the dangers: http://www.itfseafarers.org/itf-

study.cfm

W) Night & day sleep services: http://www.entandallergy.com/nad/

X) Exxon Valdez oil spill trustee council: http://www.entandallergy.com/nad/

http://aviationknowledge.wikidot.com/



http://www.abc.net.au/news/stories/2010/04/15/2873531.htm

http://www.hse.gov.uk/humanfactors/topics/fatigue.htm

http://www.hse.gov.uk/humanfactors/topics/fatigue.htm

http://controls.engin.umich.edu/wiki/index.php/Learning_and_analyzing_Bayesian_networks_with_Genieetworks_with_Genie

http://controls.engin.umich.edu/wiki/index.php/Learning_and_analyzing_Bayesian_networks_with_Genieetworks_with_Genie

http://www.marinebuzz.com/

http://www.marinebuzz.com/2008/05/07/how-seafarers-can-overcome-fatigue/



http://www.pilotfriend.com/training/flight_training/human/drugs_etc.htm

http://www.pilotfriend.com/training/flight_training/human/drugs_etc.htm

http://hfcc.dot.gov/ofm/

http://www.deir.qld.gov.au/workplace/safety/topics/index2.htm

http://www.deir.qld.gov.au/workplace/subjects/fatigue/about/index.htm

http://en.wikipedia.org/wiki/Human_factors

http://en.wikipedia.org/wiki/Human_reliability

http://www.imo.org/humanelement/mainframe.asp?topic_id=286

http://handicapsupport.com/Physical-%20Disability/Fatigue/

http://handicapsupport.com/Physical-%20Disability/Fatigue/

http://en.wikipedia.org/wiki/Fatigue_%28medical%29

http://encyclopedia2.thefreedictionary.com/Fatigue


http://www.healthcentral.com/sleep-disorders/napping-192940-5.html

http://www.healthcentral.com/sleep-disorders/napping-192940-5.html

http://www.pilotfriend.com/aeromed/medical/fatigue1.html

http://aviationknowledge.wikidot.com/


http://human-factors.arc.nasa.gov/zteam/fcp/FCP.pubs.html

http://human-factors.arc.nasa.gov/zteam/fcp/FCP.pubs.html

http://en.wikipedia.org/wiki/Stress_%28biology%29

http://www2.ulaval.ca/en/home.html

http://www.itfseafarers.org/itf-study.cfm

http://www.itfseafarers.org/itf-study.cfm

http://www.entandallergy.com/nad/

http://www.entandallergy.com/nad/



Y) Norwegian Maritime Directorate: http://www.sjofartsdir.no/no/

Z) Endyoursleepdeprivation.com: http://www.end-your-sleep-deprivation.com/

AA) Wikipedia- HFACS:

http://en.wikipedia.org/wiki/Human_Factors_Analysis_and_Classification_Sy

stem

BB) Fatigue-Healthline.com: http://www.healthline.com/

CC) The full wiki-Fatigue:

http://misc.thefullwiki.org/Fatigue_%28physical%29

DD) Wikipedia - Circadian Rhythms:

http://en.wikipedia.org/wiki/Circadian_rhythm

EE) Smart motorist-drivers fatigue: http://www.smartmotorist.com/

FF) Petri Myllymäki-Bayesian Networks : http://www.cs.helsinki.fi/u/myllymak/

http://www.sjofartsdir.no/no/

http://www.end-your-sleep-deprivation.com/

http://en.wikipedia.org/wiki/Human_Factors_Analysis_and_Classification_System

http://en.wikipedia.org/wiki/Human_Factors_Analysis_and_Classification_System

http://www.healthline.com/

http://misc.thefullwiki.org/Fatigue_%28physical%29

http://en.wikipedia.org/wiki/Circadian_rhythm

http://www.smartmotorist.com/

http://www.cs.helsinki.fi/u/myllymak/

a bayesian network application for the ...a bayesian network application for the prediction of human...

Documents